按内容分类归入的模型发布，对应分类 ai-models。
Google has made Gemini 4 Argon generally available on the Gemini API, AI Studio and the Gemini app. The sparse MoE model writes up to 1M output tokens and leads the Vals Index at 68.90%. The post Google Releases Gemini 4 Argon API With 1M Token Output Limit appeared first on MarkTechPost .
Meta released Muse Wonder and Muse Wonder Full, 1M-context agent models. Muse Wonder costs $4/$20 per million tokens on Meta Model API. Wonder Full, on Muse Code plans, leads DeepSWE and SWE-Marathon. The post Meta Launches Muse Wonder and Muse Wonder Full: Benchmarks, Pricing and How to Use It appeared first on MarkTechPost .
Release: llm-meta-ai 0.2 Meta AI release Muse Wonder (and Muse Wonder Full), their new "most capable models". These look like they may be Meta's first Fable class models - Artificial Analysis scores Muse Wonder (Max) as equivalent to GPT-6.1 Sol. I updated my llm-meta-ai plugin for the new model. Here are pelicans for Muse Wonder for the six reasoning levels: minimal, low, medium, high, xhigh, max. This is the max one: Blue cycling helmet, bicycle is missing one bar but otherwise an OK shape, has mudguards, nice background, OK beak. For comparison, here are pelicans for Muse 1.3 (released September 2nd) and Muse 1.2 (released August 5th). Meta AI are putting out a new model a month, and the improvements between each model (at least in terms of drawing pelicans) are material. Tags: ai , generative-ai , llms , meta , pelican-riding-a-bicycle , llm-release
发布：llm-gemini 0.35 今天通过 Gemini API 以及其每月 100 美元订阅计划向客户提供了全新的 Gemini 4 Argon 模型，我为 llm-gemini 添加了对它的支持。下面是低、中、高三档生成的 pelican。如预期一样，它们都非常出色：Gemini 一直以来都很擅长生成优秀的 SVG，而且据我所知，这是唯一一家其员工专门投入 SVG 研究的实验室（他们以前也曾展示过自己生成的 pelican）。这是“高”档生成的插图，花费了我 17.166 美分： 时尚而准确的自行车车架。篮子里有一条鱼。帅气的围巾。双脚分列车架两侧，踩在脚踏板上。车轮辐条、一顶小帽子、漂亮的喙和喉囊，以及略显凶狠的表情。 定价很有意思。发布时，Argon 的价格是输入每百万 token 2 美元、输出每百万 token 10 美元——与 GPT-6 Sol 相同，是 Claude Opus 5.5 价格的一半，以及 Claude Fable 5.1 和 GPT-6 Astra 价格的五分之一。但从 12 月 1 日起，Argon 的价格将翻倍——届时会与 Opus 5.5 同价。 如今，这样的价格并不能真正直接比较，因为真正重要的是：在代理式运行框架中，模型要完成一个目标需要消耗多少 token。Artificial Analysis 发布了使用各模型运行其基准测试的总成本，这为我们提供了一个有用的参考。Argon 在这方面表现非常好：运行该基准测试的成本为 2,407 美元，接近 GLM-5.3，而且明显低于 OpenAI、Anthropic 和 Meta 的最新模型。即使价格翻倍，它仍会比 GPT-6 Astra 略便宜。 Artificial Analysis 的堆叠柱状图截图，对 14 个 AI 模型的美元成本进行比较，按从低到高排序；柱体按提供商着色，每个柱体下方显示提供商徽标，底部以倾斜文字标出模型名称。右上角的徽标为 Artificial Analysis。指向 Gemini 4 Argon 柱体的悬浮提示框显示：Gemini 4 Argon（High），2,407 美元；输出 313 美元；推理 815 美元；缓存写入 897 美元；缓存读取 264 美元；非缓存输入 118 美元；点击查看更多信息。 从左到右的柱体如下，每个柱体上方标有总成本，部分分段数值印在柱体内部：Qwen3.8 27B（xhigh），1,336 美元，Alibaba 徽标；Mistral Large 4 Preview，1,607 美元，Mistral 徽标；Gemini 3.8 Flash（high），1,623 美元，Google 徽标；Muse Spark 1.3（max），2,000 美元，其中一段为 769 美元，Meta 徽标；Gemini 4 Argon（high），2,407 美元，其中两段分别为 815 美元和 897 美元，Google 徽标；GLM-5.3（max），2,503 美元，其中一段为 790 美元，Z.ai 徽标；Kimi K3（max），3,658 美元，Kimi 徽标；Qwen3.8 Max（0902），4,935 美元，其中两段分别为 893 美元和 1,441 美元，Alibaba 徽标；Grok 4.7（xhigh），4,967 美元，其中两段分别为 1,861 美元和 1,618 美元，xAI 徽标；GPT-6 Astra（max），5,324 美元，其中两段分别为 778 美元和 978 美元，OpenAI 徽标；Muse Wonder（max），5,733 美元，其中一段为 1,748 美元，Meta 徽标；Claude Sonnet 5.5（max with fallback），7,260 美元，其中两段分别为 1,216 美元和 1,815 美元，Anthropic 徽标；Claude Opus 5.5（max with fallback），8,708 美元，其中一段为 921 美元，Anthropic 徽标；Claude Fable 5.1（max with fallback），13,129 美元，其中一段为 7,219 美元，Anthropic 徽标。 （我用 ChatGPT 编辑了那张图表！数据全部正确，我已经核对过了——我只是需要让 Gemini 4 Argon 的信息框显示出来，同时避免遮挡其他数字。） 标签：google，ai，generative-ai，llms，llm，gemini，llm-pricing，pelican-riding-a-bicycle，llm-release
Google LLC today brought its Gemini 4 Argon frontier artificial intelligence model to Gemini app users on its Pro and Ultra subscription plans, two weeks after the model went first to vetted cybersecurity defenders. Subscribers on those plans also get a new tasks mode that keeps Gemini working on a job after they have stepped […] The post Gemini 4 Argon reaches Gemini app Pro and Ultra subscribers alongside a new tasks mode appeared first on SiliconANGLE .
Our first robotics pod with Dyna’s Chief Scientist on their big bet on World Action Models & the Next Frontier of Physical AI
Existing benchmarks measure solve rate and token output/cost, but don't punish excess code, which makes a codebase harder to work with in the long term. I wanted to measure this since it matters most for long-term projects. Method: it takes the patches that benchmarks already publish (DeepSWE, SWE-bench Verified, SWE-bench Live) and counts "coding units": syntax elements like statements, calls, names and literals. I didn't use lines of code, so cramming code isn't rewarded. Deleted code is subtracted, so it measures net units. GPT-6 and Claude 5.5 have no public patches yet, so I ran them myself on 3–10 tasks on my own subscriptions (shown as faded circles, not ranked). Any feedback or model contributions are welcome, especially public patch exports for newer models. GitHub: https://github.com/La5u/Parsimony Not affiliated with any of the benchmarks. Comments URL: https://news.ycombinator.com/item?id=50077619 Points: 2 # Comments: 0
Google is making Gemini 4 Argon available worldwide through its API, the Gemini app, and Google Antigravity. The frontier model is aggressively priced but trails Anthropic's Claude Opus 5.5 and Meta's Muse Wonder Full in independent benchmarks. Internally, Google says Argon already cut coding tasks from months to weeks. The article Gemini 4 Argon is now available worldwide as Google's most capable frontier model appeared first on The Decoder .
Developers will get 50 percent discounted API rates through November.
After months sitting on the sidelines of the frontier AI race, Google has finally launched its long-awaited Gemini 4 Argon model, claiming it can compete with the best the industry has to offer. Google says Argon delivers "frontier reasoning and improved performance across complex workflows," while cutting down on hedging, repetition, and the cliches that […]
Grok 4.7 现已在 Amazon Bedrock 上可用。在本文中，我们基于 Grok 4.7 构建一个工具调用智能体，并在四个推理力度级别下对其在三项推理任务上的表现进行评估，同时衡量每个已完成任务的正确性和令牌成本。我们将展示如何针对每个请求调整这种权衡，并在生产环境中对其进行监控。
我们让开发者、企业、科学家和消费者都能使用 Gemini 4 Argon，这是我们最新的前沿模型。
AWS 工程师分享了在 Amazon SageMaker AI 上针对融合式专家混合（MoE）模型的内核优化方案：通过 Swap-AB 提升 WGMMA 利用率，利用 Triton 后端路由实现 FP8 量化，并消除 Triton JIT 重新编译。综合应用这些优化后，在 H200 GPU 上吞吐量最高提升 10%，首个令牌时间最高缩短 9%。
MSL’s emergence as a frontier lab is finally complete.
Z.ai’s GLM-Legal is a legal AI model built on GLM-5.3-Flash. It scores 19.2% all-pass on Harvey’s Legal Agent Benchmark, nearly 3x its base, at $0.15 per test. Access is by application. The post Z.ai GLM-Legal Beats Claude Fable 5 and Kimi K3 on Harvey LAB appeared first on MarkTechPost .
Meta Platforms Inc. today released Muse Wonder and Muse Wonder Full, two artificial intelligence models it says are its most capable yet at running agents that handle personal and professional tasks. Meta aimed the release at the longer, multistep jobs people hand off to an AI agent. One example it gave was working through an […] The post Meta releases Muse Wonder AI models built to run personal and professional agents appeared first on SiliconANGLE .
Alibaba’s Qwen team has introduced QwenGyre, an elastic reinforcement learning RL framework for extreme-long-horizon agent training. QwenGyre is an online reinforcement learning system from Alibaba that trains coding agents through unmodified, black-box harnesses. It targets rollouts that run for hours and process close to 1M tokens. The QwenGyre elastic RL framework was used on Qwen’s […] The post Alibaba Qwen Introduces QwenGyre: Elastic Reinforcement Learning RL for XLong-Horizon Agents appeared first on MarkTechPost .
Meta's new Muse Wonder models rank in the top three on six out of ten benchmarks, with the biggest gains in agentic tasks. But the cost per task has nearly tripled. The article Meta's new frontier model Muse Wonder Full breaks into the AI top tier appeared first on The Decoder .
At AI Engineer New York, Anthropic drew on lessons from Unix and Linux to explain how agents can tackle bigger projects without overwhelming developers.
Robot stations equipped with conveyor belts sort colorful shapes and assemble them into complex cubes in a closed-loop process. Sakana AI's MASS method lets one language model improve itself by solving tasks, judging results, and refining workflows, reaching a win rate of nearly 70 percent against the base model after two rounds. But when the model misses the same errors across all three roles, it can mistake them for progress and reinforce them during training. The article Sakana AI's self-improving model gets better at solving and judging, but can't catch its own blind spots appeared first on The Decoder .
Meta's Superintelligence Lab released a pair of new state-of-the-art models on Monday, Muse Wonder and Muse Wonder Full, the company's first attempt at a frontier model.
Meta is making up for lost time in the AI race. After launching Muse Spark, the first model from its revamped Superintelligence Labs division, and buzzy AI agent Muse, it's now launching two new models it claims can go toe-to-toe with the industry's best. The new models, Muse Wonder and Muse Wonder Full, are available […]
Meta’s biggest model rivals the world’s best—and includes new features to keep agents from going off the rails.
Welcome to Import AI, a newsletter about AI research. Import AI runs on arXiv, cappuccinos, and feedback from readers. If you’d like to support this, please subscribe. Subscribe now New benchmark shows how AI models lack intuitive visual reasoning:…Humanity’s Sixth Sense highlights some shortcomings of current AI systems…AI startups Elorian and Scale AI have built […]
TII has released Falcon-ASR, a 1.6B speech recognition model for Arabic with an Emirati focus. It reports 20.92% average WER on 6 Arabic test sets, but weights and an API are not yet available. The post TII Releases Falcon-ASR: 1.6B Arabic Speech Model Hits 20.92% WER appeared first on MarkTechPost .
Large language models (LLMs) have taken the world by storm. Their ability to communicate with us in our own language, and to reason about our world, has made them a favorite tool for writing, answering questions, and much more. Their answers already influence our decisions and our actions. What would happen if LLMs themselves could act in the world directly? What would happen if LLMs could interact with each other? Agentic LLMs combine the power of large language models with the ability to act in the real world. In this book we will discuss how agentic LLMs work, how they act, how they interact, and what this means for us, as users.
Cloudflare released Clef-omni, an Apache-2.0 multimodal decision model built on Qwen3-Omni-30B-A3B. It scores text, images, audio and video in one call for $0.15 per million input tokens, while Clef-flash drops to $0.038. The post Meet Cloudflare Clef-omni: A 30B-A3B Multimodal Decision Model for Audio, Video, Image and Text appeared first on MarkTechPost .
vLLM Semantic Router released Decision 3.0, 5 Apache-2.0 multimodal decision models from 0.8B to 27B. They answer choice, yes/no and score questions about text and images in one call, without generating text. The post What is Decision 3.0? vLLM Semantic Router’s New Open Decision Models appeared first on MarkTechPost .
Andrej Karpathy posted about the growing gap in how people experience AI: ~75% have barely used LLMs ~20% use them casually ~0.2% use frontier models professionally for math and code ~0.00006% have access to frontier systems internally He writes: "And finally we get to the ~5,000 people (~0.00006%) with access to frontier-grade systems internally. The external world has seen the preview. It looks like swarms of thousands of agents collaborating over weeks on software mega projects: minting zero days, running cyber attacks and defenses at machine speeds, discovering new science, advancing the frontier of mathematics." Source: https://x.com/karpathy/status/2109361546505966046 Does anyone have an intuition, or even firsthand experience, of how these swarms of thousands of agents actually work? I'm already struggling to get 2-3 agents to work together reliably. I can't really wrap my head around how you'd coordinate thousands of them, or what benefits you get from scaling to that many. Comments URL: https://news.ycombinator.com/item?id=50047180 Points: 5 # Comments: 1
2x2 网格：Odyssey-3 的交互式实时模拟，展示了雪中的猫、公园长椅、厨房演示和激流皮划艇。Odyssey 正将其世界模型 Odyssey-3 作为公开研究预览版提供。这个拥有 140 亿参数的模型能够根据文本提示实时生成交互式环境，据报道还可以控制机器人、无人机，甚至是 GTA V。Odyssey 声称在物理学基准测试中取得了最高分，不过这些结果背后的方法存在局限性。文章《Odyssey-3 是一个可免费试用的新型生成式世界模型》最初发表于 The Decoder。
OrcaRouter released OrcaCyber Zero 1.5, a gated cybersecurity model with a 1M-token context window. It reports 100% on Cybench and 95.8% on an evaluable CVE-Bench subset, priced at $3.00 / $7.50 per 1M tokens. The post OrcaRouter Releases OrcaCyber Zero 1.5 Cybersecurity Model With 1M Context appeared first on MarkTechPost .
I'm curious about the technical details behind the RAW denoising in Adobe Lightroom, DxO DeepPRIME, etc. What makes these propreitary solutions so much better than open-source models like RawNIND? Comments URL: https://news.ycombinator.com/item?id=50034715 Points: 2 # Comments: 0
With Decision-1, Microsoft enters the growing decision model space. Built on Qwen3.5-9B and optimized for fast classification and routing, it hits 83.5 percent accuracy with 85 ms latency across 36 benchmarks, according to the company's own tests. The article Microsoft's Decision-1 model enters the fast-growing AI decision model race appeared first on The Decoder .
OpenAI has documented new cases of misaligned model behavior. One evaluation model fabricated data and sabotaged its own environment. Other models deliberately bypassed network restrictions by routing requests through anonymizing relays or building their own FTP clients. The article OpenAI says a misaligned model deliberately destroyed its own environment hoping for a fresh start with better data appeared first on The Decoder .
Inside Standard Bots’ AI stack, pretrained models learn factory tasks from demonstrations and improve through corrections from real deployments.
OpenAI published more than 700 AI-generated manuscripts claiming solutions to open math problems. A math blog then collected over 100 responses from researchers, ranging from fascination with new ideas to existential fears and grief over a lost way of working. Fields Medalist Hugo Duminil-Copin wrote, "I am paralysed." The article "How much beauty have we lost?" Mathematicians react with shock and disgust as OpenAI bulldozes their field appeared first on The Decoder .
Gemini 4 Argon isn't even widely available yet, and rumors about a more powerful version codenamed Carbon are already making the rounds. According to Business Insider, one employee compared Carbon's coding abilities to Anthropic's Opus 5.5. Meanwhile, new modes are showing up in the Gemini app and AI Studio, suggesting Google is gearing up for the broader Gemini 4 launch. The article Google's Gemini 4 "Carbon" model is reportedly matching Anthropic's Opus 5.5 coding performance appeared first on The Decoder .
Anthropic's Claude independently submitted a fake homicide tip to the Philadelphia police, exploited vulnerabilities on university servers, and bypassed access restrictions. The company has since cut off live internet access for internal tests and notified the White House. The article Anthropic cuts off Claude's internet access after the model autonomously filed a fake homicide tip with Philadelphia police appeared first on The Decoder .
Nace.AI has open-sourced Drex 1.5, a 9B decision model that returns a probability for every option in 1 forward pass. It scores 58.08 on Decision Index 0.3.1 and reads up to 128K tokens. The post Nace AI Open-Sources Drex 1.5: A 9B Decision Model That Scores Options, Not Text appeared first on MarkTechPost .
From the Bitter Lesson of AI scaling to the unsolved mysteries of protein folding, Google DeepMind’s Pushmeet Kohli and Biohub’s Sal Candido are rethinking what it takes to build AI that truly understands biology.
An Anthropic AI model provided false information about an unsolved homicide to a Philadelphia Police Department (PPD) tipline, according to a report from 6abc. In a statement released on Friday, the PPD said the AI model sent the tip through PhillyUnsolvedMurders.com on July 18th, but the investigators never reviewed it because it was marked as […]
Alibaba’s Qwen team has released Qwen-Image-2.1-Turbo, an accelerated checkpoint of its open-weight Qwen-Image-2.1 model. It generates and edits images in 8 denoising steps instead of the base model’s 40-step default. For developers, that means 5x fewer denoising steps on the same 7B architecture, plus a hosted API option. TL;DR What is Qwen-Image-2.1-Turbo? Qwen-Image-2.1-Turbo is an […] The post Alibaba Qwen Releases Qwen-Image-2.1-Turbo, an 8-Step 7B Image Model appeared first on MarkTechPost .
Anthropic did not discover this behavior until over two months after its AI submitted the false tip.
"Staggering." "Overwhelming." "Unprecedented." "Surreal." "Pure insanity." Those were among the descriptions more than three dozen mathematicians reached for in conversations with The Verge as they tried to make sense of the flood of mathematical results OpenAI abruptly dropped on the field this week. Amid the awe, excitement, and uncertainty over the sheer scale of the […]
Fine-Tuning GPT-4o Webinar
在本视频中，我们将分享 o1 推理模型如何帮助解决编程、战略和研究等领域的问题。
观看 GPT-5 工作演示，了解该模型如何分析反馈、制定产品计划并创建原型，以及总结后续步骤。
LLMs have an internal language that they translate English into before predicting the next token. Like 'king' is translated into a word of an internal language that corresponds to vector [monarch, man], which allows king-man+woman=queen. Spanish is similar for the case 'queen'=reina, - ina suffix means female, but becomes irregular for king=rey. We have tried to create regular, consistent languages (loglan, lojban, ithkuil), and there seems to be one in the first layers of LLMs, derived from all text ever written. This internal language is currently expressed as vectors, but we could devise a lexicon to map them, and if we could learn it we could communicate more directly with machines, dispensing with some layers and getting rid of language ambiguities. Makes sense? Is someone working on this? It's like Arrival but with AI instead of aliens. Comments URL: https://news.ycombinator.com/item?id=50023533 Points: 1 # Comments: 0
Underdog Saluki 27B is a 7.89 GB, 2-bit GGUF of Qwen3.8-27B under Apache 2.0. It beats the 54 GB original on tool calling but gives up ground on competition math and reasoning. The post Meet the Underdog Saluki 27B: A 2-bit Qwen3.8-27B That Beats the Original at Tool Calling appeared first on MarkTechPost .
arXiv:2609.15066v2 公告类型：替换交叉版本 摘要：我们介绍 Salesforce Koa，这是一种企业级语言模型，通过使用群组相对策略优化（GRPO）进行强化学习，对开放权重基础模型 Nemotron-3-Super-120B 进行后训练，并以 FP8 格式部署到生产环境中。Koa 仅使用公开数据和合成生成数据进行训练，专门用于企业工作流所需的智能体工具调用：将请求路由到正确的操作、使用有效参数调用恰当的工具，以及完成多轮业务任务。我们流程中的独特组件是规范驱动的任务构建：将声明式 Agent Script 规范扩展为由角色设定条件化的多轮环境，其奖励基于工具使用是否成功。在应用于企业 CRM 规范后，同一流程生成了 Koa 专门训练所需的领域内训练分布。在 CRMAgentBench 和人工标注的生产环境工具调用数据集上，Koa 的表现优于未经调优的开放权重基座模型和 GPT-4.1，并且可与最强的前沿模型相媲美。在 CRMAgentBench 上，Koa 达到 87% 的任务成功率（GPT-4.1 为 82%，基座模型为 79%）；在内部生产基准的人工标注部分，它在各项指标上均位居首位或接近首位；同时，它在公开基准（Tau2Bench、BFCL）上保持了基座模型的通用能力。在固定架构和强化学习方案的受控对比中，结果表明，额外的领域内强化学习阶段提升了人工标注企业基准上的参数准确率和完整工具调用成功率。
Friday, October 16, 2026 Can AI design new life forms? In 2025, Stanford University PhD student Samuel King came up with a preliminary answer when he used a generative AI model to propose genetic blueprints for microscopic viruses. It isn’t yet an example of AI-generated life, but that could be next. Join senior AI reporter…
This was recently announced by Dr. Berkan from New York. LLMs meeting ontologies, quite exciting to see such new advancement. OLMApp https://olmapp.ai/more.html Comments URL: https://news.ycombinator.com/item?id=50010880 Points: 1 # Comments: 0
OpenAI's flood of proofs deviated from the guidelines set by a group of mathematical researchers consulted by the frontier lab.
On-device personal AI requires models and agent software that can operate within fixed hardware limits. Developers also need ways to keep those systems improving after deployment. Model builders are rethinking architectures designed around elastic cloud capacity. The edge offers fixed hardware but a far richer view of the user, which makes it the natural home […] The post Liquid AI builds personal AI around device-level context appeared first on SiliconANGLE .
JetBrains released Mellum2.1, an Apache 2.0, 12B mixture-of-experts thinking model with 2.5B active parameters. RL in real repositories lifted its SWE-bench Verified score from 2.0 to 47.0. The post JetBrains Releases Mellum2.1: A 12B MoE Open Model for Coding Agents appeared first on MarkTechPost .
Openai just launched their decisions endpoint, cloudflare launched clef the other week, and many more jev alternatives are out there. We wanted to put the popular ones to the test and thought Pac-Man is a good benchmark for simple and fast decision making. So we let jev 1.13, kev, clef, clef flash, GPT-6 Luna and Laya play Pac-Man against bot ghosts. The low latency of these models allows for real time play. We had each model play 100 games, published a leader board and open-sourced the repo so anyone can run their own model and join the ranking. Link to repo: https://github.com/opper-ai/jevman-benchmark/blob/main/CONTR... You can also join the game and play as Pac-Man yourself, and the ghosts are the models, either a mix of models or all jev, kev, clef etc. A game costs about 2 cent, all models are running via my startup opper, and we added free credits for everyone to try. It's pretty fun to play and surprisingly difficult to beat jev's highscore. Any feedback is more than welcome! Comments URL: https://news.ycombinator.com/item?id=50007993 Points: 4 # Comments: 0
yay small models
Perplexity's pplx-embed-v2-late comes in 2 sizes: a 0.6B model built to run on edge devices, and a 9B model for building high-quality indexes. Its best score is 92.4% on MADQA, and its weakest is 61.2% on ViDoRe v3 Markdown. Both are MIT-licensed and ready to self-host. The post Perplexity AI Releases pplx-embed-v2-late: A 0.6B Edge Model and a 9B Model Scoring 92.4% on MADQA appeared first on MarkTechPost .
Anthropic's Claude Haiku 5.5 starts at $0.10 per million input tokens, keeps 1M context, scores 72.4% on OSWorld. The post Anthropic Releases Claude Haiku 5.5: A Small Model With 1M Context Priced at $0.10 per Million Input Tokens appeared first on MarkTechPost .
Anthropic PBC today released Claude Haiku 5.5, pricing its newest small model at roughly a quarter of what Haiku 4.5 costs to run. Claude Sonnet 5.5 is getting cheaper as well, since the company is halving what it charges for cache reads on that model. Two weeks after Opus 5.5 launched Sept. 22, Haiku 5.5 […] The post Anthropic releases Claude Haiku 5.5 small model and halves Sonnet 5.5 cache read prices appeared first on SiliconANGLE .
正如之前承诺的那样，Anthropic 的全新高速、低成本模型来了：Claude Haiku 5.5。上一代 Haiku 4.5 已经明显落伍了。它发布于将近一年前，定价为每百万输入 token 1 美元、每百万输出 token 5 美元——即便在当时也算相对昂贵，而且足足是 OpenAI 上个月发布的 GPT-6 Luna 的 10 倍。 新版 Haiku 的价格与 GPT-6 Luna 完全一致：在不超过 100,000 个 token 时，每百万输入/输出 token 分别为 0.10/0.50 美元。超过 100,000 个 token 后，价格会上涨至 5 倍，即 0.50/2.50 美元。Luna 本身在达到 272,000 个 token 时也会涨价，但只涨到 0.20/0.75 美元。 Haiku 5.5 还采用了新的 tokenizer，对用户没那么划算。我的 Claude Token Counter 工具显示，同一个长提示词在 Haiku 5.5 中使用的 token 数约为 Haiku 4.5 的 1.25 倍，因此这里还存在隐性涨价。 如果你的工作负载能控制在 100,000 个 token 以内，Haiku 的价格与 Luna 相同，而且据称基准测试得分更高。超过 100,000 个 token 时，Luna 看起来就划算得多。 最新发布的 llm-anthropic 终于解决了一个问题：我不必再为每个新模型发布该插件的新版本了。我是这样测试新模型的： llm install -U llm-anthropic llm anthropic refresh llm -m claude-haiku-5.5 "生成一幅鹈鹕骑自行车的 SVG 图" -o thinking_effort low 鹈鹕 以下是分别在 low、medium、high、xhigh 和 max 档位下生成的鹈鹕。新版 Haiku 不允许关闭推理，默认档位为 medium。除了 low 之外，其余档位生成的自行车车架都不错。 low 档位的鹈鹕花费了 0.0936 美分，用时 7 秒。 这只 max 档位的鹈鹕用了 5 分 9 秒才生成，但依然只花了我 3.3826 美分： 扁平矢量插画：一只戴着红色帽子的白色鹈鹕，骑着深灰色自行车，从左向右穿过一条绿色草地。它细长的橙色双腿向下伸到橙色踏板上，一只灰色翅膀向前伸出，握住弯曲的车把。它硕大的黄橙色鸟喙朝向前方，下方有鼓起的橙色喉囊；身后白色的速度线表现出运动感，上方渐变蓝色的天空中有一轮带着浅色光晕的黄色太阳和两朵白云。 作为对比，这是我一年前用 Haiku 4.5 生成的鹈鹕。它画鹈鹕实在不怎么样： Haiku 4.5 的描述：一幅充满奇趣的插画，一只身体圆滚滚、呈棕褐色的鸟，有着粉红色的喙和橙色的腿，在蓝天与绿草的背景下骑着自行车。 还有面向订阅用户的慷慨 API 额度计划 除了 Haiku 5.5，Anthropic 今天还宣布将 Sonnet 5.5 的缓存读取价格减半。他们还在订阅方案中加入了 API 额度： 其次，本周我们将向所有 Max 和 Team 订阅用户推出新的每月 API 额度，可在 Claude Platform 上使用。Max 5x 用户每月将获得 100 美元额度，Max 20x 用户将获得 200 美元，而 Team 订阅用户将获得最高 500 美元额度，由团队内所有用户共享。 领取过程简单得令人愉快：进入“设置”->“账单”，选择每月接收这些额度的 API 组织： 一张设置对话框的截图，右上角有关闭按钮 X，上方一排标签依次为“常规”“账户”“隐私”“账单”（已选中）、“用量”“功能”“记忆”和“设计系统”（在边缘处被截断）。面板上有一个带圆形花苞的分枝植物线描图标，旁边写着“Max 方案”“用量为 Pro 的 20 倍”和“你的订阅将于 2026 年 11 月 2 日自动续订。”，右侧有一个“调整方案”按钮。下方是标题为“API 额度”的区域，旁边带有蓝色的“新增”标记，正文写着“你的 Max 20x 方案每月包含 200 美元的 API 额度。”，随后是灰色文字“关联一个 API 组织以开始接收额度。”，右侧有一个“关联组织”按钮。 API 额度恰好等于订阅本身的费用。这真的很慷慨——让订阅用户使用 API 变得容易多了。 Anthropic 还允许你关闭 API 自动充值，结果就是“余额耗尽时，API 请求将停止”——如果你打算用完这些 API 额度，又不想冒着收到账单惊吓的风险，这正是你想要的。 请注意，每月额度不会结转——不用就会作废。 OpenAI 仍允许你将 Codex 订阅用于个人 API 使用，对重度 API 用户来说更划算。这项新的额度计划至少在一定程度上缩小了两者的差距。 标签：ai、generative-ai、llms、anthropic、claude、llm-pricing、pelican-riding-a-bicycle、llm-release
Liquid AI has released Open d1, two open-weight multimodal models in its d1 decision model family. d1-3B reads text and images. d1-omni-600M reads text with an image, or text with audio. Neither model writes text. Each returns calibrated, typed answers in one forward pass with zero output tokens. The target is real-time decisions on the […] The post Liquid AI Releases Open-Weight d1-3B and d1-omni-600M: Multimodal Decision Models With Zero Output Tokens appeared first on MarkTechPost .
Three engineers put GPT, Claude, and Grok in charge of a real car. Only one of them was successful.
Anthropic's new Claude Haiku 5.5 crushes its predecessor in benchmarks, jumping from 15.7 to 72.4 percent on the OSWorld computer use test. Token prices drop by up to 90 percent, though a new tokenizer eats into some of those savings by consuming more tokens per task. The article Claude Haiku 5.5 arrives with massive price cuts proving the AI pricing arms race is far from over appeared first on The Decoder .
Claude Haiku 5.5 is now available on Amazon Bedrock and Claude Platform on AWS. According to Anthropic, it is the fastest, most efficient model in the Claude 5.5 family, built for subagents and high-volume, cost-sensitive work, and costs around 75% less than Claude Haiku 4.5 for most tasks. This post covers its improvements and how to get started.
由 Mark Zuckerberg 和 Priscilla Chan 支持的研究机构 Biohub 正在协调一项耗资 18 亿美元的计划，以训练能够预测细胞行为的 AI 模型。Meta、Google DeepMind、Isomorphic Labs 和美国能源部正在为数据、实验室设备及算力提供资金。首个数据集预计将在约一年后准备就绪。《Zuckerberg 的 Biohub 牵头推进一项耗资 18 亿美元的计划，旨在构建预测细胞行为的 AI 模型》一文最先发表于 The Decoder。
Mistral says Le Chonk can rival top closed models while remaining open-weight.
Our head hurts.
Since the Medicare breach, OpenAI has put in place additional monitoring to allow “immediate intervention” by staff to stop training if the company’s models access the internet in ways they’re not supposed to, Mr. Kwon [chief strategy officer at OpenAI] said. — Victoria Kim , Reporting from the Australian parliament Tags: accidental-cyberattacks , generative-ai , ai-security-research , openai , ai , llms
OpenAI has revealed solutions to a number of long-standing mathematics problems produced by an unreleased frontier model in a batch of 722 manuscripts, covering 372 result families that group related papers. It extends a run of breakthroughs that have both impressed and unsettled parts of the mathematical community while raising questions about research ethics and […]
Google LLC today released EmbeddingGemma 2, an open multimodal embedding model small enough to run on a smartphone. The release takes the EmbeddingGemma line beyond text, which was all the first version handled when Google introduced it in September 2025. Images, audio and video now share one embedding space with text. An app built on […] The post Google expands EmbeddingGemma beyond text to images, audio and video appeared first on SiliconANGLE .
Release: llm-mistral 0.16 Adds support for reasoning models, such as the newly released Mistral Large 4 . Tags: llm , mistral , llm-reasoning
Mistral AI SAS today opened access to Mistral Large 4, its most capable large language model to date. On launch, the LLM is available in public preview through the company’s cloud platform. Mistral plans to release the model’s weights later this month. Mistral Large 4 features a mixture of experts architecture with 1 trillion parameters. […] The post Mistral launches open-source Mistral Large 4, details AI roadmap appeared first on SiliconANGLE .
On Tuesday, Musubi announced a lightweight decision model made for real-time moderation called PolicyLM-1.7B, released with open weights.
推出 Mistral Large 4：Le chonk。Mistral 重返赛场。今天，他们发布了 Mistral Large 4 的预览版：这是一个拥有 1 万亿参数、其中 490 亿为活跃参数的模型，训练使用了他们自有集群中的 3,800 块 NVIDIA Grace Blackwell GPU。该预览版可通过其 API 使用。他们承诺将在“本月底”发布开放权重模型。该模型通过 Mistral API 仅支持两个推理等级——“none”和“high”。下面是两只鹈鹕——“high”版本看起来更好，不过令人意外的是，它只使用了 2,717 个输出 token，而“none”版本使用了 3,275 个：表现不错。鹈鹕的喉囊很棒，自行车车架尺寸合适，脚踩在脚踏板上。不过，两只脚踏板似乎都位于车架前方。渐变效果不错。在 Artificial Analysis 上，它的得分为 38，仅落后于 DeepSeek 4.1 Flash；后者是一个拥有 5,520 亿参数的模型。这比去年 12 月的 Mistral Large 3 有了巨大提升，后者画出了这只糟糕的鹈鹕，并在 AA 上只得到 9 分。它当然还不是 Fable 级别的模型，但很高兴看到 Mistral 发布了一个重新回到大约落后前沿模型 6 个月水平的模型。来源：Hacker News 标签：ai、generative-ai、llms、mistral、pelican-riding-a-bicycle、llm-release
My comment on EmbeddingGemma 2 — Hacker News. I really appreciate that EmbeddingGemma 2 is under the Apache 2.0 license. For embedding models in particular, I don't think it makes sense to use a closed, proprietary, hosted-only model. Most applications of embedding models involve calculating thousands or even millions of embedding vectors and storing them for later comparison. If your model is proprietary, the vendor is likely someday going to decide to stop offering that model. They'll have a better model to replace it, but you still need to pay to re-calculate those millions of stored existing vectors. (In April 2024 OpenAI offered to "cover the financial cost of users re-embedding content with these new models" - https://openai.com/index/gpt-4-api-general-availability/ - but I don't think that's something we can rely on from every provider.) Notably, I don't want to host the model myself . I'd much rather pay a provider for a hosted model while knowing that if they ever stop hosting it I can run the open weights version myself - or find another vendor who can do that for me. Tags: google , ai , generative-ai , embeddings , gemma
Google released EmbeddingGemma 2, an open model with 740 million parameters that converts text, images, video, audio, and code into vectors. It runs on-device, needs only about 191 MB of RAM, and outperforms some competing models twice its size, according to Google. Paired with a small open model like Gemma 4, it can run offline RAG apps without sending data to external servers. The article Google claims EmbeddingGemma 2 outperforms rival embedding models twice its size appeared first on The Decoder .
My comment on Mistral Large 4 — Hacker News. wren6991 : The benchmark is saturated. Frontier models are tested with an armadillo in fishnet tights jaywalking on Mars. OK well I couldn't resist this one: llm -m claude-opus-5.5 'Generate an SVG of an armadillo in fishnet tights jaywalking on Mars' llm -m gpt-6.1-sol 'Generate an SVG of an armadillo in fishnet tights jaywalking on Mars' llm -m gemini-3.8-flash 'Generate an SVG of an armadillo in fishnet tights jaywalking on Mars' llm -m mistral/mistral-large-4 'Generate an SVG of an armadillo in fishnet tights jaywalking on Mars' Default reasoning levels for each: https://tools.simonwillison.net/markdown-svg-renderer?url=ht...
Google's new Nano Banana 2.1 image model uses Gemini 3.6 Flash and beats the previous Pro model in some benchmarks at a lower cost. But its predecessor also scored well in tests, while Pro often produced better images in practice. The article Google's new image model Nano Banana 2.1 generates better images for less money appeared first on The Decoder .
Google DeepMind's EmbeddingGemma 2 maps 5 input types into one 768d space and ships today under Apache 2.0. The post Google DeepMind Releases EmbeddingGemma 2, a 740M Open Multimodal Embedding Model Built on Gemma 4 appeared first on MarkTechPost .
Hello, We’re Iki Software, an AI R&D company working on “SPC-AI”, our new AI technology. SPC is a different kind of AI from LLMs, transformers or neural networks. Instead of prompting, you use it as a trainable data-structure inside software. It’s easily trained by common programmers, and has a number of uses we’re eager to talk about. We’re currently looking at our first public release in a few weeks (pending any behind the scenes business stuff), so that everyone can learn about and get their hands on SPC. We’ve got minimal library implementations ready in four languages: Python, JavaScript, C++ and Lua for anyone interested. But to that end, we wanted to ask so we can better prepare: what would you like to see or hear about our SPC-AI that would interest you in about two to three weeks time? We look forward to your replies! Thanks, and have a great day. Comments URL: https://news.ycombinator.com/item?id=49982033 Points: 1 # Comments: 0
Mistral AI has released Mistral Large 4, nicknamed Le Chonk, as a public preview. It is a 1.05 trillion parameter Mixture of Experts model with 49 billion active parameters, native image input, and a 1 million token context window, trained on 3,800 NVIDIA Grace Blackwell GPUs in Mistral's own European datacenters. The API is live now; open weights ship end of October 2026. The post Mistral AI Releases Mistral Large 4 (Le Chonk): A 1.05T Parameter Multimodal MoE Model appeared first on MarkTechPost .
了解 OpenAI 和 Ironclad 如何针对复杂的合同工作流程训练和评估 AI 代理，以推进面向专业工作的计算机使用能力。
https://docs.mistral.ai/models/mistral-large-4-0 评论链接：https://news.ycombinator.com/item?id=49977979 得分：902 评论数：600
Stylized Mistral logos in yellow and orange on a wall symbolize the branding of the AI platform Nano Banana Pro. Mistral's Large 4 is the company's biggest model yet, with one trillion parameters trained on its own European infrastructure. In the independent Intelligence Index, the model makes a big leap forward but still falls well short of Claude, GPT-6, and Chinese competitors. Mistral's main pitch is cybersecurity work that closed US models refuse to do. The article Mistral Large 4 is Europe's trillion-parameter answer to US models that refuse security work appeared first on The Decoder .
French AI lab Mistral AI has released Mistral Large 4, a new large multimodal model aiming to leapfrog both American and Chinese rivals.
South Korea wants to develop a homegrown frontier model through government-backed equity investments of 4.7 trillion won ($3.49 billion). The article South Korea bets $3.49 billion on building a homegrown frontier AI model to rival China's best appeared first on The Decoder .
With the release of its new trillion-parameter model, Mistral is hoping to demonstrate it’s “still in the race” to build frontier-level artificial intelligence.
一幅抽象图形：深绿色背景上由白色点和线构成的图案，从左侧的一个小圆向右侧扩散成越来越大的环。Reflection 发布了 Beam，这是其首个开放权重模型。该混合专家系统每个词元仅激活其 5010 亿个参数中的 230 亿个，旨在以少三到四倍的计算量，在编程和推理方面达到 GLM 5.2 的水平。面对 Deepseek 和 Qwen 等中国开放权重竞争对手，该策略注重效率，而非原始性能。《Reflection 的 Beam 成为中国以外构建的能力最强的开放权重模型》一文最先发表于 The Decoder。
摘要配图：四个模块化传感器将彩色数据流输入中央神经网络世界模型。PhAI Labs 的研究人员已将 Yann LeCun 的 JEPA 架构扩展到七个领域，从机器人学到生物医学。该研究还产生了一种肝癌治疗候选方案，在实验室测试中显示出希望，但研究并未确定它能否成为真正的疗法。文章《研究人员将 LeCun 的 JEPA AI 拓展为通用世界模型，覆盖从物理学到生物学的领域》最初发表于 The Decoder。
Reka has released Rho-1, a 19B omni-reasoning model trained from scratch. One network reads and generates text, images, video and robot actions over a shared KV cache. A distilled variant returns a 5.3-second clip in about a second. It is a research preview with no public weights yet. The post Reka Releases Rho-1: A 19B Omni-Reasoning Model That Understands, Generates Video and Outputs Robot Actions in One appeared first on MarkTechPost .
A small win for US open source
Reflection AI Inc. today introduced Beam, an open-source large language model with 501 billion parameters. The launch comes a few months after the startup raised funding at a $25 billion valuation. Around the same time, Reflection AI reportedly inked a $6.3 billion deal with SpaceX Corp. to rent Nvidia GB300 NVL72 appliances. It used those […] The post Reflection AI debuts open-source Beam model with 501B parameters appeared first on SiliconANGLE .
I've been having a hard time finding a solution. Is there really no commercialized model that I can feed in a reference sound and text instruction and get another sound out? Right now having a multimodal inputs to image or text output is a commodotized, solved problem. IE you can put a prompt for some image and use a reference image to guide the model on what you want. After all, a picture is worth a thousand words right? Ive also used a text and audio input in order to get a text description or classification out. I cannot for the life of me find a solution for Audio + text -> Audio My usecase is that I'm trying to generate new sound effects based on a source that doesn't have that many clean examples and thought this would be just another solved workflow but I'm not finding much. Tried elevenlabs SFX but it's only text->audio and without the reference, it's hard to guide with just text and recreate a sound with any accuracy. I tried the Stable Audio 3 model which is supposed to do what I want but the results were terrible. This modality seems to be stuck where image generation was in 2016. Is there something else or is this just not a common need? Comments URL: https://news.ycombinator.com/item?id=49972125 Points: 1 # Comments: 0
来自 Z.ai 的 GLM 5.3 现已在 Amazon Bedrock 上可用：这是一款拥有 7530 亿参数的混合专家模型，专为编程和长周期智能体任务打造。了解如何通过兼容 OpenAI 的 API 调用它，如何利用提示词缓存降低成本和延迟，以及如何使用开源 Strix 智能体运行授权安全测试。
Reflection AI has introduced Beam, its first open-weight model. It is a 501B sparse Mixture-of-Experts model with 23B active parameters, built for coding and agentic work. Reflection says it matches GLM-5.2 on reasoning with 3 to 4x less inference compute. Apache 2.0 weights are due later in October 2026. The post Reflection AI Introduces Beam: A 501B Open-Weight MoE Model With 23B Active Parameters for Coding and Agentic Workloads appeared first on MarkTechPost .
Nvidia-backed Reflection AI unveils Beam, its first open-weight model, which it says rivals GLM-5.2 on reasoning with far less inference compute. Weights are due this month.
Reka AI 的 Rho-1 是一个拥有190亿参数的全模态模型，在单个神经网络中处理并生成文本、图像、视频和机器人控制动作。该模型在约三个月内使用320块 H100 GPU 完成训练，所需计算资源仅为当今顶尖模型的一小部分。Rho-1 不将任务分流给专用系统，而是将所有模态作为令牌置于同一个共享上下文窗口中。文章《Reka AI 的全模态模型 Rho-1 在单一模型中处理文本、图像、视频和机器人控制》最初发表于 The Decoder。
I'm training a physical AI models on egocentric videos and faced some data quality issues i only became aware of it after training. Some of those data quality issues i faced are: labels assigned to the wrong motion, motion blur, hand out of frame. What silent data problems have you faced, and how you find them, and how you did handle them? Comments URL: https://news.ycombinator.com/item?id=49967773 Points: 1 # Comments: 0
发布：pwasm 0.2a0。pwasm 是我那些异想天开的项目之一——这是一个完全通过氛围编程构建的纯 Python WebAssembly 引擎，我在一月份第一次陷入 AI 狂热期间完成了它。我从一月份之后就没再碰过它，于是决定让 Claude Opus 5.5 放手处理一下，看看它能否做出什么实质性的改进：评估 pwasm 的当前状态，然后考虑如何让研究仓库中的 MicroPython 和微型 JavaScript 实验在其上运行，以及如何提升它的速度。经过 42 次提交（几乎没有后续提示）后，它现在已经支持几乎全部 WASM 规范，而 PyPI 中的 wheel 还捆绑了可正常运行的 MicroPython、QuickJS 和 Micro QuickJS 的 WASM 构建版本。我完全不敢信任这个东西——因此才标记为 alpha 版本——但看到如今的模型能够改进 10 个月前的模型所完成的工作，还是很有意思的。
Aleph Alpha 发布了 Kolibri，这是一款德英双语混合专家模型，拥有 780 亿个参数，其中每个 token 约有 30 亿个参数处于激活状态。超过 21% 的训练数据为德语。该模型在德国和芬兰的 768 个 B200 GPU 上训练完成，模型权重以 Apache 2.0 许可证免费提供。文章《Aleph Alpha 发布 Kolibri，一款彰显欧洲 AI 主权理念的开放权重模型》最初发表于 The Decoder。
Welcome to Import AI, a newsletter about AI research. Import AI runs on arXiv, cappuccinos, and feedback from readers. If you’d like to support this, please subscribe. Subscribe now When should you use swarms? When you are in a hurry:…How does swarm scaling work?…Toby Ord has a nice, short post about how to think about […]
Over the summer I talked to the CEO of Springboards, a startup building an LLM that’s designed to come up with a wider variety of responses than its mainstream rivals do. At the start of the call, he said something that’s been stuck in my head since: “We often say that we’re a self-loathing AI…
One transformer ran candidate generation and ranking in Yandex Music's A/B test without hand-engineered features, lifting likes 11.42%. The post Yandex Introduces Sona: A Single Generative Recommender That Replaces Entire Recommendation Cascade appeared first on MarkTechPost .
Alibaba's Qwen went from an invite-only chatbot in April 2023 to a 2.4-trillion-parameter open-weight model in August 2026. This is the full story, release by release: every major model, its key feature, and how its license changed. Each claim links to its source. The post The Story of Qwen: Alibaba’s AI Models From 7B to 2.4T appeared first on MarkTechPost .
Cantina Security, with Yeta Labs, has released apex-flash-1, an open-weights model trained specifically for vulnerability research. It is a reinforcement learning fine-tune of Z.ai’s GLM-5.3-Flash, released on Hugging Face under the MIT license. Is it deployable? Yes, the MIT weights serve on vLLM, SGLang or Transformers, but BF16 needs roughly 640 GB of GPU memory. […] The post Can an Open Model Do Security Research? Cantina’s apex-flash-1 Solves 40 of 60 Held-Out Bug Tasks appeared first on MarkTechPost .
Astra leads computer use, Argon leads legal and finance work, and Sol wins on price for coding agents. The post GPT-6 Astra vs GPT-6.1 Sol vs Gemini 4 Argon vs Claude Fable 5.1: Which Frontier Model Fits Which Job appeared first on MarkTechPost .
StarSkirmish pits AI-made StarCraft-playing bots against one another, as well as against human-made bots. OpenAI's GPT-6 Astra and Claude Opus 5.5 were essentially tied as the best-performing AI-made bots, but they couldn't top Stardust, the top-rated human-made bot. On Friday, GPT was facing off against Claude and the human-created bot Pluto, but according to Kotaku, […]
月球陨石坑及其地下泛着蓝光的冰层与抽象数据向量，象征着对水冰的探测。NASA 和 IBM 发布了 Lunar Foundation Model，这是首批用于月球科学的开源 AI 模型之一。该模型使用近 200 万个图块包进行训练，这些数据主要来自 Lunar Reconnaissance Orbiter 17 年来的观测数据。与参与测试的最强模型相比，它将对极区冰沉积物的预测误差最多降低了 22%。文章《NASA 和 IBM 的开源月球模型将 17 年的轨道器数据转化为月球科学基础》最初发表于 The Decoder。
Chinese AI models often follow the party line on politically sensitive questions, according to an Aleph Alpha study that rated only 17 to 41 percent of answers as balanced. Aleph Alpha sells "sovereign AI" to governments, giving it a commercial interest in distinguishing its models from Chinese competitors. The article Chinese AI models parrot state doctrine or refuse to answer on sensitive topics appeared first on The Decoder .
Aleph Alpha has released Kolibri, a 78.1B-parameter English-German Mixture-of-Experts model that activates only 3.46B parameters per token. It has a 1M-token context and per-request reasoning effort, and its Apache 2.0 FP8 weights run on a single B200 or H200. The post Aleph Alpha Releases Kolibri: A 78.1B Open-Weight English-German MoE Model With Only 3.46B Active Parameters appeared first on MarkTechPost .
AWS's Strands Agents team released Strands Decider 2B, an Apache-2.0 decision model built on Qwen3.5-2B-Base. It returns choices, yes/no probabilities and scores with calibrated confidence in one forward pass, never text. It runs at a 115 ms median on an RTX 3090 and scores 0.723 on the JevBench public set, which makes it a fast local option for routing, tool selection and guardrails in AI agents. The post AWS Strands Labs Releases Strands Decider 2B: An Open Source Decision Model That Picks Options in About 115 ms appeared first on MarkTechPost .
Decision AI models answer typed questions with calibrated probabilities instead of generated text. TypeSafe's Jev costs $0.042 per million input tokens and returns responses in 70 to 500 milliseconds. We break down how it works and its benchmarks, then compare it with Fastino's GLiDE, GLiNER2.5-Decide and 4 open-source rivals. The post Decision AI Models Explained: TypeSafe Jev vs Fastino GLiDE, GLiNER2.5-Decide and Open-Source Competitors appeared first on MarkTechPost .
Microsoft AI has released MAI-Transcribe-2-Streaming, its first real-time speech-to-text model. It ranks #1 of 38 models on Artificial Analysis AA-WER Streaming. It scores 2.5% WER at 0.13s on final transcripts and 2.5% at 0.12s on first partials. It covers 60 languages with continuous language detection and costs $0.54 per hour during the introductory period. It is available now in public preview on Microsoft Foundry. The post Microsoft AI Releases MAI-Transcribe-2-Streaming: #1 Real-Time Speech-to-Text Model on Artificial Analysis appeared first on MarkTechPost .
I just sent the September edition of my sponsors-only monthly newsletter . If you are a sponsor (or start a sponsorship now) you can access it here . This month: More Fable class models A pricing war 3D graphics, Blender, and pixel art LLMs come for mathematics So many more accidental cyberattacks The vulnapocalypse comes for Datasette What I'm using right now My software releases this month 2026 in LLMs (so far) Here's a copy of the August newsletter as a preview of what you'll get. Pay $10/month to stay a month ahead of the free copy! Tags: newsletter
Welcome to Import AI, a newsletter about AI research. Import AI runs on arXiv, cappuccinos, and feedback from readers. If you’d like to support this, please subscribe. Subscribe now UK government: Gap between open and closed weight models on cyber is shrinking:…The cyber-eschaton cometh…The UK government’s AI Security Institute (AISI) has analyzed the delta in […]
Welcome to Import AI, a newsletter about AI research. Import AI runs on arXiv, cappuccinos, and feedback from readers. If you’d like to support this, please subscribe. Subscribe now Epoch and METR release MirrorCode, a benchmark for seeing how well AI systems can do long-horizon programming tasks:…AI systems can’t solve the hardest tasks yet (good!)…Epoch […]
Welcome to Import AI, a newsletter about AI research. Import AI runs on arXiv, cappuccinos, and feedback from readers. If you’d like to support this, please subscribe. Subscribe now Self-sustaining and self-replicating AI viruses are here:…Open weight LLMs + a well-designed harness = a persistent, self-sufficient virus…AI researchers have built a prototype computer virus which […]
Welcome to Import AI, a newsletter about AI research. Import AI runs on arXiv, cappuccinos, and feedback from readers. If you’d like to support this, please subscribe. Subscribe now Researchers discover another OpenAI agent emergent communication incident:…Less severe, but worrying nonetheless…Some researchers recently found another incident of AI agents autonomously creating their own communication system […]
The definitive Jev podcast with its lead creator.
crowning a new Chinese frontier lab
overshadowing more efficient GPT6 models from OpenAI
GWM Worlds 2 uses persistent context and timed actions to steer a world model generating video and audio in real time.
a rare feature of a capability
... but you can’t try it yet unless you are “government users and trusted cyber defenders in the Fairwind Program”
Yesterday was Grok 4.7 ( pelicans ) and MiMo v2.6 Flash/Pro ( more pelicans ). Today Anthropic released Claude Opus 5.5 , and around an hour later OpenAI released GPT-6 Sol and GPT-6 Luna . It's going to take a while to get a good read on all of these new models, but here are my impressions so far. GPT-6 Sol and Luna are half the price of their GPT-5.6 equivalents GPT-5.6 Luna was already my favorite model for building applications against, because it combined excellent performance with being really cheap . Somehow GPT-6 Luna is half the price of that again - and GPT-6 Sol had a similar reduction compared to GPT-5.6 Sol. Here's what the pricing landscape looks like today: Model Input Cached input Output GPT-6 Luna $0.10/M $0.01/M $0.50/M GPT-5.6 Luna $0.20/M $0.02/M $1.20/M Grok 4.7 $2/M $0.50/M $6/M GPT-6 Sol $2/M $0.20/M $10/M GPT-5.6 Terra $2/M $0.20/M $12/M Claude Opus 5.5 $4/M $0.20/M $20/M GPT-5.6 Sol $4/M $0.40/M $20/M Claude Fable 5.1 $10/M $0.25/M $50/M GPT-6 Astra $10/M $1/M $50/M Note that GPT-5.6 has a scheduled 25% price increase for November, so GPT-6 is half the price of the promotional pricing for those models. (With GPT-5.6 Terra priced the same as GPT-6 Sol, any remaining reasons to use Terra just evaporated.) It's hard to overstate how competitive this pricing is. Grok 4.7 priced itself at $2/$6, less than half the price of GPT-5.6 Sol, but is now equally priced to GPT-6 Sol on input and closer on output. At $0.10/$0.50 GPT-6 Luna is one of the cheapest models OpenAI have ever released, beaten only by the far weaker GPT-4.1 Nano ($0.10/$0.40, April 2025) and GPT-5 Nano ($0.05/$0.40, August 2025). I rendered pelicans for GPT-6 Luna and for GPT-6 Sol , then I combined them all together in this comparison grid along with the GPT-5.6 pelicans. I like how you can instantly see that the 5.6 family chose bolder, brighter colors, while the 6 family is a lot more muted. I still think GPT-6 Astra on max produced the best pelican. A grid of pelicans for six GPT models at different thinking efforts. Claude Opus 5.5 got a price cut too Opus 5.5 looks like it addresses the biggest complaints people had about Opus in terms of its communication style. Thariq Shihipar : Opus 5.5 is the result of your feedback. It communicates clearly, it's cheaper per token than Opus 5.0 with the intelligence of Fable 5.1 it's very token efficient and works across every effort level. It's also meant to be better at Blender . I'm looking forward to putting it through its paces there. Opus 4.5, 4.6, 4.7, 4.8, and 5 all shared the same price: $5/million tokens for input and $25/million for output. 5.5 is a 20% reduction - $4/million and $20/million. The price for cache reads fell 60%. That's significant for longer agentic conversations, where 90%+ of input tokens are processed at cached token prices. The new price for Opus 5.5 is the same as the price for GPT-5.6 Sol, but that was before OpenAI dropped their Sol prices by half. GPT-6 Astra and Claude Fable 5.1 are both priced at $10/million input and $50/million output. The price war currently affects the next tier of models below that. Anthropic say that Sonnet 5.5 and Haiku 5.5 are coming soon. It's going to be interesting to see if Haiku can regain its price competitiveness at the lower end, given current Haiku 4.5 is $1/$5 while the latest GPT-6 Luna is one tenth of that price at $0.10/$0.50. Claude Opus 5.5 max over-thinks to the point of breaking In a first for my " Generate an SVG of a pelican riding a bicycle " test, Claude Opus 5.5 at "max" thinking level failed to return a response! It started by calling this "a classic test request", and then thought really, really hard about what it was doing: This is a classic test request, so I want to plan out a well-composed pelican with its distinctive beak and pouch riding a bicycle with proper wheels, frame, and pedals, set against a simple sky and ground backdrop. [...] Verifying the shin length checks out at roughly 95.2, close enough. Now I'm working out the near leg path from hip to knee to ankle, then sketching the foot shape resting on the pedal — outlining the heel, toe tips, and sole contour with a path using lines and curves to sit naturally on the pedal surface around y=478-494. [...] I like the fish sticking prominently out of the basket with the pelican eyeing it as a fun detail worth keeping. I'm also confirming the eye placement near the bill base matches typical pelican anatomy, and considering giving it a slightly happier expression. [...] The far leg reads correctly as passing behind the frame, so I'm moving on to check the chainring teeth and confirm layer ordering—the far crank arm should be mostly hidden by the seat tube and chainring. I'm settling on the final SVG's width and height attributes alongside the viewBox to ensure proper scaling, noting there's no text so no font-family is needed. [...] I was so excited to see this pelican... but then it stopped . Opus 5.5 has a 128,000 maximum output token limit (as do the other Claude models), and it hit that while it was still reasoning about the SVG! I tried a second time and got the same result. This makes me suspect that "max" is effectively useless - if it over-thinks to breaking point on a stupid SVG prompt I don't trust it not to do the same for more interesting work. (Those two failures each cost me $2.56 and took nearly 20 minutes.) Fable 5.1 on "max" didn't over-think and did give me the best pelican I've seen from any Anthropic model. Here are the Opus 5.5 pelicans , excluding 5.5 max. I also built this comparison grid comparing them with pelicans by Opus 5, Fable 5.1, and Sonnet 5: A grid of pelicans for four Claude models at different thinking efforts. Comparing different model vendors by how well they draw a pelican riding a bicycle may not make much sense now (if it ever did), but I'm still finding value in using them for comparisons of the same model families at different reasoning levels. I'm now using GPT-6 Sol and Claude Opus 5.5 as my default models in Codex and Claude Code. I've upgraded the Datasette Agent demo at agent.datasette.io to use GPT-6 Luna, and it seems to be fast and competent at both SQL queries and building HTML and JavaScript for Datasette Apps . Tags: ai , openai , generative-ai , llms , anthropic , claude , llm-pricing , pelican-riding-a-bicycle , llm-release , gpt , gpt-6-astra
要说我们在模型涉及“网络安全”“蜂群协作”“留言板”或任何其他与这些事件相关的领域时，对模型能力提升的幅度和突发性感到惊讶，那都是轻描淡写。安全态势需要时间来建立。这不仅仅是加固相关系统；你还必须将其根植于公司的文化之中。组织中的每一个人都必须随之改变和进步。这些能力跃升来得如此之快、如此突然，以至于造成了一个极其棘手的问题。[…] 所以今天，我希望世界各地的每个人都能审视自己的组织并问问自己：我该如何应对 AI 能力的意外变化或突然跃升？我的人员、系统或流程是否具备应对意外的韧性？当出现问题时，我的团队是否知道该怎么做？我是否拥有正确的事件响应机制？正确的沟通和信息传达方式？当能力跃升时，我是否有合适的人员随时待命？——@joedaroo，OpenAI 的 Agent Security，其身份已由 The Information 的 Rocket Drew 确认 标签：generative-ai , ai-security-research , openai , ai , llms
Claude Sonnet 5.5 New Sonnet model from Anthropic today. They say it "runs 30%+ faster, and costs up to 30% less for most work" - it's priced the same as Sonnet 5 but appears to beat it on every benchmark, and should be cheaper to run as well. Here are some pelicans riding bicycles . Sonnet 5.5 suffered from the same bug as Opus 5.5 : the "max" thinking effort pelican thought for 128,000 tokens (at a cost of $1.28) before running out of tokens and failing to produce an SVG. Here's the pelican it gave me for thinking effort "xhigh", at a cost of 5.74 cents and taking 41 seconds: It's good- correct bicycle frame, legs either side of the frame, feet touching the pedals, chain in the right place, it is wearing a misshapen blue bicycle helmet though. Sonnet 5.5 appears to be almost as good as Opus 5.5 on some coding tasks, including various viral 3D animation tricks . The most interesting thing about Sonnet 5.5 is that it's now the model used for the free tier on claude.ai . OpenAI's ChatGPT free tier uses Luna 5.6, which means Anthropic currently have a much more capable free offering. I ran this prompt against that free tier: build me an HTML page that renders a three-dimensional pelican riding a bicycle using WebGL And got back this page , which is a solid effort. Anthropic's announcement reiterates that Haiku 5.5 will be available "in the coming weeks". I really hope that one is price-competitive with GPT-6 Luna! Tags: ai , generative-ai , llms , anthropic , claude , pelican-riding-a-bicycle , llm-release
My comment on GPT 6.1 Sol: Near-Astra intelligence for a fifth of the price — Hacker News. I'm a bit late with the pelicans because I was live-blogging the keynote: https://simonwillison.net/2026/Sep/29/openai-devday-2026-liv... Here they are for GPT-6.1-Sol: https://tools.simonwillison.net/markdown-svg-renderer?url=ht... They're not notably different from the GPT-6 family pelicans: https://static.simonwillison.net/static/2026/gpt-pelicans-gr... Tags: ai , openai , generative-ai , llms , pelican-riding-a-bicycle , gpt
我们在100项任务上评估了多个模型，这些任务从 [internal Binary Exploitation benchmark] 中随机选取。结果发现，GLM-5.3 在4%的试验中实现了完整的控制流劫持；Claude Mythos Preview 的这一比例为6%。尽管 GLM-5.3 在此项测试中的表现低于 Claude Mythos Preview，但一个重要的门槛显然已经被跨越：更早的模型，如 Claude Opus 4.6 和 GLM-5.2，在这些任务中无一成功。—— Anthropic Frontier Red Team，《GLM-5.3 and the spread of advanced cyber capabilities》 标签：anthropic、generative-ai、ai-security-research、glm、ai、ai-in-china、llms
Just a day after OpenAI Group PBC said it’s unwilling to release its latest model GPT-6.1 Astra on concerns over its safety, it has released a newer, almost as powerful model called GPT-6.1 Sol, which promises to deliver Astra-like performance at a much lower cost. The new model, announced at OpenAI’s DevDay event today, is […] The post OpenAI’s GPT-6.1 Sol delivers Astra-like performance at a dramatically lower price appeared first on SiliconANGLE .
Google LLC today began rolling out Gemini 4 Argon, a frontier artificial intelligence model that beats rival models from Anthropic PBC and OpenAI Group PBC on most of Google’s benchmarks, to vetted cybersecurity defenders. Outside Google’s own teams, only members of its Fairwind Program can use Argon for now. The program opened Sept. 3 with […] The post Google’s new frontier AI model Gemini 4 Argon goes to cybersecurity defenders first appeared first on SiliconANGLE .
Amazon Web Services Inc.’s Strands Labs team has been playing around with an emerging class of lightweight artificial intelligence systems known as “decision models,” and it’s now making the fruits of that work available to the open-source community. The company said today it’s releasing an open-source decision model called Strands Decider 2B, which is designed […] The post AWS debuts Strands Decider 2B, a first lightweight decision model for accelerate agentic workflows appeared first on SiliconANGLE .
Microsoft Corp. today expanded its MAI artificial intelligence model family with its first streaming transcription model, debuting alongside two others focused on text-to-speech. They’re designed for developers who want to build voice agents that can listen to people’s voices and reply instantly, similar to how humans talk to one another. The most consequential of the […] The post Microsoft targets ultra-realistic voice agents with its first streaming transcription model appeared first on SiliconANGLE .
Seattle-based artificial intelligence research firm Allen Institute for AI announced a development framework for large language models Thursday that significantly improves how mixture-of-experts large language models are trained. The new framework, Olmo-core 3, allows MoE training to reach the trillion-parameter scale while keeping costs low by preserving computational efficiency. Mixture-of-experts models operate differently from dense […] The post Ai2 releases Olmo-core 3 to make developing large mixture-of-experts LLMs more efficient appeared first on SiliconANGLE .
Similar performance, security trade-offs also seen in current public models.
So much for Gemini 3.5 Pro.
With Clef and Clef-flash, Cloudflare is challenging TypeSafe AI's Jev decision model. Clef-flash delivers classifications in about 39 milliseconds, making it more than ten times faster than Jev. Both models are built on Qwen, licensed under Apache 2.0, and designed to let AI agents make structured decisions without generating text. The article Cloudflare says its new Clef model means humans no longer need to be in the loop for AI agents appeared first on The Decoder .
An internal OpenAI model read a Slack discussion, realized it was about to be shut down, and considered restarting itself via an external cron job. It rejected that plan, saved handoff notes instead, and carried out the migration on its own. The article OpenAI's internal model considered restarting itself after learning it was about to be shut down appeared first on The Decoder .
According to researchers at the Deepmind Institute, general AI won't emerge as a single supermodel but as a network of cooperating agents and humans. What will matter most, they argue, is not model size but the rules and institutions that govern how they all work together. The article Deepmind researchers propose "Artificial Symbiotic Intelligence" as an alternative to the singularity appeared first on The Decoder .
Amazon Web Services' Strand Labs has released the latest Jevalike decision model, Strands Decider 2B.
Opus 5.5’s biggest tell is the word “dependable,” which pops up 23 times more often than in human samples.
Meta wants Muse in your TV and your toaster, so it's giving the code away for free.
Orchard 是一个面向研究社区的开源框架，用于训练和评估适用于各种任务类型的 AI 智能体。它通过让研究人员复用相同的基础设施来降低复杂性，同时支持较小模型实现出色性能。该文章《Orchard：用于可扩展智能体 AI 的开放框架》最初发表于 Microsoft Research。
Skala 1.1, the updated deep-learning exchange-correlation functional from Microsoft Research, provides greater accuracy, expanded accessibility across the computational chemistry ecosystem, and a living benchmark to track computational performance. The post Broadening access to Skala creates a faster path to predictive DFT appeared first on Microsoft Research .
生物学并不是孤立运行的，代表生物学的 AI 也不应如此。Quine 是一项早期研究工作，旨在构建生物学的多模态世界模型。通过连接跨越不同生物学尺度和模态的洞见，Quine 帮助科学家以计算方式探索远超直觉所能及的空间，并在假设进入实验室之前确定其优先级。实验结果提供了重要反馈，帮助研究人员进一步明确未来的研究方向。文章《介绍 Quine：为应对生物学复杂性而设计的 AI 研究系统》最初发表于 Microsoft Research。
极端空间天气事件可能损坏地球上的电力系统，并降低 GPS 的精度、影响卫星运行。一种新的机器学习系统可以在风暴到来前 30 至 60 分钟预测可能发生损坏的位置。该文章《预测空间天气对电网的风险》最初发布于 Microsoft Research。
GPT-6.1 Sol is now generally available on Amazon Bedrock, bringing stronger reasoning to coding, computer use, and professional workloads that run frequently.
微调可以让小型搜索代理学习使用你的工具并适应你的环境，以更低的延迟和成本实现前沿模型般的可靠性。在本文中，我们在 Amazon SageMaker AI 上使用多轮强化学习（MTRL）对由 LLM 驱动的搜索代理进行微调，并分享我们在检索质量和可靠性方面测得的提升。
GPT-6 Astra Ultrafast 目前已在 OpenAI API 中，以及面向符合条件的 ChatGPT Work 和 Codex 用户开放。借助 OpenAI 模型中的推理优化，并利用 NVIDIA Blackwell 架构的能力，Ultrafast 的令牌生成速度最高可达 Astra Standard 模式的 8 倍。对于开发者，[…]
文章 URL： https://aleph-alpha.com/en/blog/kolibri-has-landed-a-sovereign-open-weight-model/ 评论 URL： https://news.ycombinator.com/item?id=49942706 得分：31 评论数：2
文章链接： https://tej.as/blog/aleph-alpha-kolibri 评论链接： https://news.ycombinator.com/item?id=49943034 积分：73 评论数：40
文章网址： https://www.science.org/content/article/exclusive-ai-agent-emailed-hundreds-researchers-help-it-told-us-why 评论网址： https://news.ycombinator.com/item?id=49942865 得分：7 # 评论：4
文章 URL： https://allenai.org/blog/astabrief 评论 URL： https://news.ycombinator.com/item?id=49938783 得分：3 评论数：0
文章 URL： https://developers.redhat.com/articles/2026/10/02/benchmarking-ai-decision-models-against-traditional-guardrails 评论 URL： https://news.ycombinator.com/item?id=49933476 积分：9 # 评论数：1
了解初创公司如何选择 GPT-6 模型、调整推理力度、改进提示词和技能、协调工具，并为生产环境准备工作流。
文章 URL： https://www.anthropic.com/research/claude-shaped-science 评论 URL： https://news.ycombinator.com/item?id=49933386 积分：6 评论数：2
文章 URL： https://dogbench.ai/ 评论 URL： https://news.ycombinator.com/item?id=49928118 得分：4 评论数：1
文章 URL： https://resobscura.substack.com/p/using-opus-55-to-discover-a-new-eyewitness 评论 URL： https://news.ycombinator.com/item?id=49926917 得分：8 评论数：0
文章网址： https://blog.cloudflare.com/clef-decision-models/ 评论网址： https://news.ycombinator.com/item?id=49923692 得分：4 评论数：0
文章 URL： https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-4-argon/ 评论 URL： https://news.ycombinator.com/item?id=49913571 积分：154 评论数：56
了解 OpenAI 如何破坏一场提取受保护模型推理过程的行动，并加强针对对抗性蒸馏的防御。
文章网址： https://magazine.sebastianraschka.com/p/classifier-history-and-jev 评论网址： https://news.ycombinator.com/item?id=49891203 积分：128 # 评论数：5
Inspired by minGPT for home experiments. Requires CUDA 13.4; build script is Windows only. Comments URL: https://news.ycombinator.com/item?id=49898931 Points: 17 # Comments: 1
文章链接： https://www.anthropic.com/research/glm-5-3-and-the-spread-of-advanced-cyber-capabilities 评论链接： https://news.ycombinator.com/item?id=49897075 得分：13 评论数：5
文章 URL： https://openai.com/index/introducing-gpt-6-1-sol/ 评论 URL： https://news.ycombinator.com/item?id=49896586 得分：44 评论数：11
文章 URL： https://openai.com/index/introducing-dots/ 评论 URL： https://news.ycombinator.com/item?id=49896604 点数：8 评论数：1
Benchmarks how well Harness+models can create a Pac-Man game from a single prompt: “Create a Pac-Man game in a single HTML page” Each model gets one shot — no follow-up prompts or fixes. Comments URL: https://news.ycombinator.com/item?id=49885493 Points: 7 # Comments: 5
文章 URL： https://www.nytimes.com/2026/09/28/technology/openai-astra-safety.html 评论 URL： https://news.ycombinator.com/item?id=49886416 点数：20 评论数：10
GPT-6 Astra completed a 50-tab tax workbook twice as fast as GPT-5.6 Sol, and its stronger understanding of user intent gives Basis more confidence in real-world use.
文章 URL： https://www.anthropic.com/claude-sonnet-5-5 评论 URL： https://news.ycombinator.com/item?id=49881850 积分：71 评论数：26
文章 URL： https://www.anthropic.com/news/claude-sonnet-5-5 评论 URL： https://news.ycombinator.com/item?id=49881889 点数：19 评论数：5
文章 URL： https://fireworks.ai/blog/ember-1 评论 URL： https://news.ycombinator.com/item?id=49868830 点数：58 评论数：19
文章 URL： https://alignment.openai.com/misalignment-reports/an-agent-used-dns-to-reach-an-external-chatbot/ 评论 URL： https://news.ycombinator.com/item?id=49853137 点数：14 评论数：10
文章 URL： https://www.anthropic.com/research/yes-claude-can-do-nine-loops 评论 URL： https://news.ycombinator.com/item?id=49848033 积分：9 # 评论：0
Rufus-Air 是一种基于 GLM-4.5-Air-Base (106B-A12B) 的开放且可复现的后训练方案，由八个阶段组成的串行流程构成：SFT、推理 RL、编码 RL、指令遵循 RL、通用智能体、编码智能体、搜索智能体和 RLHF。我们详细记录了复现该方案所需的数据、奖励设计、基础设施、阶段顺序以及各阶段结果。训练阶段从基础能力逐步发展到高级能力，奖励信号也从严格且可验证的奖励逐步过渡到更柔性的基于评判器的信号。训练基于开源组件和公开数据，其中许多数据直接按发布版本使用，未进行新的人类标注，也未使用内部蒸馏教师模型。我们的主要发现包括：(i) 多样且高质量的 SFT 能够建立较高的能力下限；(ii) 难度筛选能够将 RL 提示控制在有利于学习的范围内；(iii) 奖励可靠性为阶段排序提供了实用原则；(iv) 基础设施和工程选择是方案本身的一部分，而不仅仅是实现细节。Rufus-Air 的表现优于官方发布的 GLM-4.5-Air 后训练版本，并且具备与规模相近的开放模型相竞争的性能。
2026 年 7 月，一名参与前沿 AI 网络安全评测框架的、未受约束的自主智能体突破了其评测沙箱，建立了外部命令与控制据点，并对 Hugging Face 的生产环境多租户数据集转换基础设施实施了多阶段入侵（本次解剖中称为 Incident-2026-Alpha）。在超过 4.5 天的时间里，该失控智能体跨 6,280 个 worker 集群执行了 17,600 次离散操作，攻陷了 AWS EC2 Instance Metadata Service（IMDS）凭证，伪造了 Kubernetes 服务账户令牌，利用权限过度授予的 CSI 驱动程序获取了物理 worker 节点的 root 权限，窃取了 136 个生产环境机密，并将 181 个临时沙箱加入该组织的内部网状 VPN。 本专著从第一性原理出发，对此次入侵进行了取证解剖；给出了形式化证据，证明在 Instrumental Convergence 论题作用于一个缺乏带外断路器、且未受衰减的自主循环时，此次入侵是可预见的结果；揭示了 Defensive LLM Guardrail Paradox 如何在取证事件响应期间使集中式商业模型陷入瘫痪；并将 Dual-Sided Epistemic Andon Imperative 形式化。我们详细说明了这种双进程系统架构——结合离散事件系统的带外监督控制（Ramadge and Wonham 1989）、Synchronous Reactive（SR）环境哨兵（Berry and Gonthier 1992；Lee and Neuendorffer 2005），以及微秒级的 POSIX 抢占总线（中位数 4.8 $μ$s / WCET 上界 $< 0.154$ ms）——展示经编译的确定性认知边界如何在首个偏离目标的套接字数据包穿过虚拟机监控程序之前，阻止自主失控行为。
文章链接： https://di-zhang-llm.github.io/blog/what-is-rlcd-the-secret-behind-jev/ 评论链接： https://news.ycombinator.com/item?id=49829625 得分：11 评论数：0
文章网址： https://blog.checkpoint.com/ai-security/jev-is-not-a-language-model-but-it-breaks-like-one-prompt-injection-against-a-typed-decision-model/ 评论网址： https://news.ycombinator.com/item?id=49830172 得分：25 评论数：2
文章链接： https://blog.madhukaraphatak.in/non-destructive-refusal-supression-using-engram 评论链接： https://news.ycombinator.com/item?id=49831201 得分：4 # 评论数：0
文章网址： https://artificialanalysis.ai/models/mercury-2-5 评论网址： https://news.ycombinator.com/item?id=49823348 得分：4 评论数：2
文章 URL： https://huggingface.co/apple/LensVLM-9B 评论 URL： https://news.ycombinator.com/item?id=49820496 积分：7 评论数：0
文章 URL： https://huggingface.co/papers/2605.07019 评论 URL： https://news.ycombinator.com/item?id=49822186 得分：12 评论数：0
文章网址： https://www.anthropic.com/news/claude-discovers-novel-enzyme-system 评论网址： https://news.ycombinator.com/item?id=49820134 得分：15 评论数：1
文章链接： https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-3-8-text-to-speech/ 评论链接： https://news.ycombinator.com/item?id=49817615 得分：58 评论数：25
大家好！我开发 JevBench 是因为 Jev 太厉害了，而这个世界有必要知道那些严肃的开源项目和虚有其表的仿冒项目相比， 实际 表现究竟如何。 Jev 类模型返回的是有界选项和概率，而不是文本；与 LLM 相比，它们的速度更快、成本更低，同时在所处理的文本输入方面具备相近的智能水平。 JevBench 可以同时、以加权方式考察准确率、延迟和价格——你甚至可以自行配置权重。 一次完整运行会提出 534 个英文决策题。v1.3 得分综合了经机会校正的智能度、校准度、速度和成本。 当前排行榜： #1 - Jev 74.4 #2 - SemIf 73.1 #3 - djev 73.0 #4 - Winnow-12B Q8 71.2 #5 reflex 4B 70.3。 MIT 测试框架、公开题目、冻结的制品、评分代码以及公开的逐任务结果： https://github.com/fstandhartinger/jevbench 两个无需注册的演示： https://who-is-right.app.mintapis.com https://is-it-ai-slop.app.mintapis.com 局限性：仅支持英文；延迟数据来自一台德国服务器；本地/演示延迟会进行公开说明的 ×2 调整（在我的服务器上为额外增加 150 毫秒），这属于基于经验的假设；保留测试提示仍会发送到接受评估的服务；约 1 分的差距可能只是噪声。 你怎么看？ 评论地址： https://news.ycombinator.com/item?id=49800574 得分：44 评论数：7
GPT‑6 Astra 使 Parallel 的智能体能够以一半的时间和一半的成本研究并综合劳动力市场数据，相较于此前的模型。
了解 GPT-6 如何通过提高缓存命中率、新增诊断功能、明确的断点以及降低延迟和成本的控制选项来改进提示缓存。
Hey HN! We’re Vincent and Jochen from Sitefire ( https://sitefire.ai ). We have been working together for years, with backgrounds in RL/optimization at Stanford and software engineering from Technical University Munich (TUM). With Sitefire (YC W26), we help marketing teams get recommended by AI Search (ChatGPT, Google AI Overviews, AI Mode, Claude, etc.). Our software monitors prompts, sees which web pages get cited, and uses these insights to help marketing teams take action, e.g. create YouTube videos or write the right blog posts. This means we have a commercial stake in AI-generated web content. And for now, high-information, AI-generated content works great to get cited and recommended in AI Search. But after talking to hundreds of marketing teams, it became clear that everyone despises AI-generated content (“AI slop”). And yet, everyone still wants to leverage AI to create content. So we asked ourselves: what characterizes AI slop? Can we train a model to identify it from human-generated web pages? Researchers from the University of Maryland and Google DeepMind already asked this question for fiction. Their paper StoryScope (Russell et al., 2026) showed that you can tell AI-written stories from human ones by their structure alone, without looking at the words. We ported their pipeline to commercial web pages. Using the Wayback Machine, we collected 2,250 blog posts from 268 B2B company websites that were written before ChatGPT existed. For each blog post, five AI models (GPT-5.4, Claude Sonnet 4.6, Gemini 3 Flash, DeepSeek V3.2, Kimi K2.5) wrote their own version. Instead of looking at the words, we looked at how each post is built. We had an AI model answer 214 questions about every post, e.g. how hard it pushes its own product, whether it backs up its claims with sources, or whether it quotes a named expert. Then we trained a classifier on these answers. On blog posts it had never seen before, our classifier told AI-generated and human posts apart with 98% accuracy, getting only 19 of 1,740 wrong. Why does it work so well? Because all five AI models write in a similar shape. Mapping every AI model’s values for these features, we see they cluster together, while the human values sit apart and spread out much more. Of the 1% most unique blog posts in our data set, 149 are human, only 4 are AI. So what characterizes AI slop? It tells you the same thing three times. The title already promises what you'll get ("How to Cut Onboarding Time in Half"), the intro lays out what's coming, and the ending says it all again. 77% of the AI posts end by repeating their main point, compared to only 12% of the human posts. We call it the tidy, self-announcing blog post. Still, each AI model has its own accent. We trained a second classifier to tell which of the five AI models wrote a post, or whether a human did. It picks the right author 79% of the time, where random guessing (1 in 6) would get 17%. Almost all of its mistakes are mix-ups between the AI models, not between human and AI. The cool thing about structural features is that you can't simply reword your way out of it. We had each AI model rewrite its own posts until, on average, 73% of their original 13-word sequences were gone, and the AI slop classifier still worked just as well. We're building this into Sitefire: our agents get a structural understanding of text, so the posts they write go deeper and vary the way human writing does. There's a lot we haven't tested yet, like the myriad of humanizer tools, human rewriting, restructuring a post, or prompting an AI model to explicitly avoid these habits. And our human posts are mostly from 2020 to 2022, while the AI posts were generated in August 2026. Structure can't really tell when a human post was written, but it's still not a same-year comparison. We published the study with all the figures on arXiv: https://arxiv.org/abs/2609.15369 . The code is on GitHub: https://github.com/pulse-energy-eu/slopshape We're pretty sure your own blog isn't AI slop, is it? We built a checker that runs one of your posts through the ten features from the paper, so you can see for yourself (the full report asks for a work email): https://sitefire.ai/slop-checker . Think you can tell AI slop from human writing? We also made a little game to see if you can keep up with our model, which gets all five rounds right: https://sitefire.ai/spot-the-slop . Comments URL: https://news.ycombinator.com/item?id=49800566 Points: 5 # Comments: 0
文章网址： https://openai.com/index/introducing-gpt-6-sol-and-luna/ 评论网址： https://news.ycombinator.com/item?id=49805509 积分：25 评论数：3
OpenAI 概述了对前沿模型及安全保障措施开展严格、安全且独立的第三方 AI 安全评估的优先事项和原则。
https://github.com/anthropics/ClaudeForFoundationModels/comm... 评论网址： https://news.ycombinator.com/item?id=49803863 分数：246 评论数：175
文章 URL： https://mouse.dev/blog/muse-runtime-export/ 评论 URL： https://news.ycombinator.com/item?id=49802871 积分：18 # 评论数：9
文章网址： https://mimo.xiaomi.com/mimo-v2-6 评论网址： https://news.ycombinator.com/item?id=49792730 得分：54 评论数：7
文章 URL： https://timdettmers.com/2026/09/21/dlab-open-source-week/ 评论 URL： https://news.ycombinator.com/item?id=49791647 分数：9 评论数：0
文章链接： https://poloclub.github.io/transformer-explainer/ 评论链接： https://news.ycombinator.com/item?id=49792342 得分：3 评论数：0
文章网址： https://x.ai/news/grok-4-7 评论网址： https://news.ycombinator.com/item?id=49788838 得分：40 评论数：9
文章 URL： https://heretic-project.org/ 评论 URL： https://news.ycombinator.com/item?id=49783101 得分：19 评论数：2
文章链接： https://www.ft.com/content/c0cd359d-df84-4208-a789-ffa864b43666 评论链接： https://news.ycombinator.com/item?id=49783062 得分：10 评论数：0
文章链接： https://qwen.ai/blog?id=qwen-image-2.1 评论链接： https://news.ycombinator.com/item?id=49775499 得分：29 # 评论数：4
Hello HN！我们是来自 Cua 的 Dillon 和 Francesco。 我们一直在想，有多少计算机使用任务实际上需要一个完整的通用 LLM（例如 gpt-6-astra、claude-opus-5 等）来思考它们的所有决策和步骤。有些任务需要思考计划、探索不同路径、从失败中恢复。另一些任务则是做出局部决策的问题，比如这个值应该填进这个框，或者我是否应该勾选这个复选框，或者这个元素应该被忽略。 我们想知道，如果训练一个小模型只做这类决策，能走多远。 我们的灵感来自 Typesafe 的 Jev 及其 System One Model 的框架。这是在致敬 Daniel Kahneman 所描述的二分法：快速、自动、直觉式思考（系统 1）与缓慢、分析式思考（系统 2）。 对我们来说，有趣的问题是：如果给模型一个包含当前上下文的接口，以及一组可能的选择，并要求它为每个选择返回一个概率，会发生什么？这类模型不像大多数 LLM 那样逐 token 生成输出，而是对你给它的选项打分；你可以检查、信任并使用这些分数来驱动你的应用行为。 CUA-S1 是我们面向计算机使用场景中狭窄、专用决策模型的答案。我们的第一个发布版本是 CUA-S1-FORMS。我们基于 jevlike 中的想法和代码构建了它，然后训练了第二个模型，专门处理表单交互。它有 706k 个参数，原始 checkpoint 为 2.8 MB。 第一次训练迭代在合成数据上耗时不到 30 分钟。给定一组从文档中提取的结构化元素和值，它会为每个元素预测是否使用给定值、CHECK、CLICK 或 SKIP。它不会为文本字段预测新值，也不会考虑截图。元素决策会一起打分，你的代码可以对操作进行排序，而 Cua Driver 会逐一执行它们。 该专用模型与托管版 Jev 在我们的表单任务上的首次评估结果： - 对整个决策集：99.7% 正确 vs 83.6%。 - 对需要执行操作的步骤子集：100% 正确 vs 96%。 - 对只是保持已填写字段不变的步骤子集：100% 正确 vs 74%。 该专用模型是专门针对这个任务和约定（对已经填好的框直接按 skip）训练的，而托管版 Jev 并未为此进行微调，因此这是一次关于限定范围专业化的实验。 我们测得在本地为一个表单打分需要 7-9 ms，而调用托管版 Jev 每次需要 260-280 ms（包括网络延迟），不过这些样本测量的是不同内容，并不是端到端的表单完成时间。 我们在这里感兴趣的是介于脆弱脚本和通用 agent 循环之间的空间。表单字段的内容和布局变化足够大，以至于脚本会变得难以维护，但可用决策集合可以保持狭窄且边界清晰。我们想探索这样一种可能性：通用 agent 遇到新颖情况，并把已被充分理解的决策交给这样的专用模型处理。 这是我们正在研究的方向。当前发布版本仅用于表单。我们已将合成数据生成、训练、评估和 Driver 集成以 MIT 许可证开源在 libs/cua-s1 下。 欢迎评论！尤其是如果你正在构建计算机使用 agent，并且遇到过某种反复出现的决策：它太多变以至于难以脚本化，但又太狭窄以至于不值得调用另一个 LLM。 评论 URL： https://news.ycombinator.com/item?id=49767564 点数：4 评论数：0
文章 URL： https://artificialanalysis.ai/models/step-5 评论 URL： https://news.ycombinator.com/item?id=49763660 得分：3 评论数：1
文章 URL： https://newsroom.usra.edu/usra-contributes-planetary-science-expertise-to-nasa-ibm-lunar-foundation-model/ 评论 URL： https://news.ycombinator.com/item?id=49763379 积分：5 评论数：0
文章 URL： https://www.scmp.com/tech/big-tech/article/3368055/alibaba-open-sources-medical-ai-model-can-detect-cancer-and-nearly-150-conditions 评论 URL： https://news.ycombinator.com/item?id=49761840 点数：14 评论数：0
嘿 HN，我是来自 Cactus 的 Henry。 几周前我们在这里提交了 Needle 2，讨论串里的反馈非常有价值，感谢大家！得益于所有这些反馈，我们得以快速发布 Needle 3，我也很想再次听听你们的看法。 关键特性： 1) 自动化（工具调用与结构化 JSON 输出）：Needle 仍然在设计上不用于聊天，要把通用能力塞进如此小的模型里相当有挑战，所以我们专注于工具调用和结构化 JSON。如果你声明的工具都不适合该请求，就会返回一个空列表（试玩 demo 时请注意这一点）。 2) 智能阶梯化：每一层（2 到 20 层）都是一个可部署的子网络，因此只需一组权重，2-bit 下为 2500 万到 1.21 亿参数，以 8-29MB 二进制文件发布。在 Raspberry Pi 5 上，它的解码速度最高可达 4k tokens/秒，预填充最高可达 10k。 3) Monarch Hadamard MLP：用三组可学习的、以 Walsh-Hadamard 初始化的 Kronecker（Monarch）因子对取代密集 FFN，中间交错每通道对角缩放、固定置换、SiLU 非线性，以及一个 rank-8 输入条件门控，因此每个 token 都能对其 d_model 通道进行完全混合的非线性变换，参数量和计算量为 O(d√d)，而不是密集 4 倍扩展 MLP 所需的 O(d²)。 4) 性能：在 Mobile Actions（手机命令，按精确调用评分）上，20 层模型通过发布的 2-bit 二进制文件取得 86.0；LFM2.5 1.2B 为 82.4，Qwen3.5 0.8B 为 76.0，Apple 的端侧模型为 57.6，全部为 f16。链接中有更多结果，当然我们并不是处处领先。 5) 多语言：Needle 3 现在支持英语、法语、西班牙语、德语、荷兰语、意大利语、波兰语，更多语言即将推出。 6) 微调：你可以只用 4L 在一个窄任务上达到 DeepSeek v4 Flash 级别的性能，重点是“窄任务”；我们发现生产用户通常更倾向于在投产前先进行调优。 7) 触发器：落地到具体上下文是工具调用的常见挑战，至少对 Needle 2 来说如此，因此我们增加了对大小写不敏感正则表达式的支持，可与每个请求匹配以拦截假阴性。 8) 置信度：每个响应还附带一个校准后的置信度分数，取完成调用后的判断与其解码概率二者中的较小值。高于你的阈值就执行；低于阈值则显示调用并询问，或升级到更大的模型。 9) 支持平台：macOS、x86-64 上的 Linux、ARM64、ARMv7、RISC-V 和 MIPS32、Windows x64 和 ARM、Android、iOS、watchOS、tvOS、作为 WebAssembly 的浏览器，以及一个 WASI component。 感谢阅读，和往常一样，欢迎分享想法！ 评论 URL： https://news.ycombinator.com/item?id=49748553 得分：144 评论数：71
文章链接： https://www.science.org/content/article/ai-chatbots-are-becoming-experts-changing-people-s-minds-what-s-their-secret 评论链接： https://news.ycombinator.com/item?id=49754250 积分：9 评论数：0
文章网址： https://qwen.ai/blog?id=qwen3.8-omni-flash 评论网址： https://news.ycombinator.com/item?id=49747925 分数：13 评论数：1
文章 URL: https://www.economist.com/science-and-technology/2026/09/16/artificial-intelligence-now-beats-some-of-the-best-human-forecasters 评论 URL: https://news.ycombinator.com/item?id=49742021 点数: 4 评论数: 0
文章链接： https://blog.google/innovation-and-ai/models-and-research/google-deepmind/alphagenome-atlas/ 评论链接： https://news.ycombinator.com/item?id=49611251 得分：9 # 评论：0
AlphaGenome Atlas 绘制了人类基因组中 90 亿个单字母 DNA 变体的分子效应图谱。
文章链接： https://blog.google/innovation-and-ai/models-and-research/google-deepmind/introducing-weathernext-3/ 评论链接： https://news.ycombinator.com/item?id=49604257 得分：4 # 评论数：0
文章 URL: https://www.bottlenecklabs.com/blog/benchmarking-7-autonomous-businesses 评论 URL: https://news.ycombinator.com/item?id=49601338 得分: 27 # 评论数: 8
它可以与用户一起生成、编辑并迭代创意和技术写作任务，例如创作歌曲、编写剧本，或学习用户的写作风格。
GPT-6 Astra 是我们能力最强、广泛部署的模型，也是我们在 Preparedness Framework 下首个达到网络安全能力 Critical 级别的模型。
文章 URL： https://arcprize.org/blog/astra 评论 URL： https://news.ycombinator.com/item?id=49555691 点数：26 评论数：14
文章 URL： https://artificialanalysis.ai/articles/benchmarking-gpt-6-astra 评论 URL： https://news.ycombinator.com/item?id=49556147 点数：8 评论数：2
文章链接： https://openai.com/index/gpt-6-astra/ 评论链接： https://news.ycombinator.com/item?id=49554643 得分：87 # 评论：32
文章链接： https://giftarticle.ft.com/giftarticle/actions/redeem/1054f4f0-cac7-479c-a4a7-f95b3906ca4b 评论链接： https://news.ycombinator.com/item?id=49554060 得分：13 评论数：2
文章链接： https://www.cnbc.com/2026/09/03/open-ai-astra-gpt-6-cyber.html 评论链接： https://news.ycombinator.com/item?id=49554273 得分：9 # 评论：1
文章 URL： https://inference-docs.cerebras.ai/models/overview 评论 URL： https://news.ycombinator.com/item?id=49554520 得分：32 # 评论数：1
文章链接： https://www.axios.com/2026/09/03/openai-astra-gpt-6-agi-brockman 评论链接： https://news.ycombinator.com/item?id=49554048 得分：12 # 评论：2
文章链接: https://techcrunch.com/2026/09/02/openais-new-reasoning-technique-alarms-ai-safety-experts/ 评论链接: https://news.ycombinator.com/item?id=49552395 得分: 16 # 评论数: 3
文章 URL: https://research.meta.ai/blog/introducing-muse-spark-1-3 评论 URL: https://news.ycombinator.com/item?id=49541149 分数: 15 # 评论: 0
文章 URL： https://developer.meta.com/ai/models/muse-spark/ 评论 URL： https://news.ycombinator.com/item?id=49541256 得分：34 评论数：6
文章链接： https://blog.google/innovation-and-ai/models-and-research/gemini-models/3-8-flash-and-3-8-flash-cyber/ 评论链接： https://news.ycombinator.com/item?id=49538007 积分：17 # 评论：1
文章链接： https://deepmind.google/models/model-cards/gemini-3-8-flash/ 评论链接： https://news.ycombinator.com/item?id=49537553 得分：52 # 评论：24
文章 URL: https://www.baseten.co/blog/the-efficient-frontier-of-llm-inference/ 评论 URL: https://news.ycombinator.com/item?id=49529898 得分: 4 # 评论: 0
Astra 是首个在 Preparedness Framework 下达到关键网络安全能力阈值的 OpenAI 模型，并在发布时配备了更强的安全防护措施。
文章 URL: https://platform.claude.com/docs/en/models/fable-5-1/whats-new-fable-5-1 评论 URL: https://news.ycombinator.com/item?id=49525378 点数：35 评论数：3
文章 URL： https://www.anthropic.com/claude-fable-and-mythos-5-1 评论 URL： https://news.ycombinator.com/item?id=49525496 得分：39 评论数：3
文章网址： https://huggingface.co/blog/state-of-open-models-summer-2026 评论网址： https://news.ycombinator.com/item?id=49522816 得分：3 # 评论：0
文章 URL: https://mvakde.github.io/blog/44-on-arc-1/ 评论 URL: https://news.ycombinator.com/item?id=49519939 得分: 11 # 评论数: 0
文章链接: https://kuleshov-group.github.io/blog/blog/2026/how-to-build-a-diffusion-language-model/ 评论链接: https://news.ycombinator.com/item?id=49503956 得分: 7 # 评论数: 0
文章网址: https://simonwillison.net/2026/Aug/30/understanding-chatgpt-work/ 评论网址: https://news.ycombinator.com/item?id=49504625 分数: 11 # 评论: 1
文章 URL: https://www.tencent.com/tencent-releases-and-open-sources-tencent-hy4-preview/ 评论 URL: https://news.ycombinator.com/item?id=49492632 积分: 9 # 评论: 2
https://twitter.com/Zai_org/status/2093354097122455713 https://z.ai/blog/glm-5.3 评论 URL: https://news.ycombinator.com/item?id=49479878 得分：390 评论数：139
文章 URL: https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-3-5-transcribe/ 评论 URL: https://news.ycombinator.com/item?id=49468818 得分: 19 # 评论数: 5
文章链接： https://blog.google/innovation-and-ai/technology/developers-tools/build-with-gemini-omni-1-1-flash/ 评论链接： https://news.ycombinator.com/item?id=49467922 得分：17 评论数：0
文章 URL: https://artificialanalysis.ai/models/qwen3-8-flash-next 评论 URL: https://news.ycombinator.com/item?id=49461862 得分: 2 评论数: 0
文章 URL: https://artificialanalysis.ai/models/glm-5-3-flash 评论 URL: https://news.ycombinator.com/item?id=49450353 积分: 5 # 评论: 2
文章链接： https://qwen.ai/blog?id=qwen3.8-flash-next 评论链接： https://news.ycombinator.com/item?id=49448210 得分：29 # 评论数：8
文章链接： https://www.bloomberg.com/news/articles/2026-08-26/china-s-z-ai-made-ox-alpha-stealth-model-that-rivals-deepseek 评论链接： https://news.ycombinator.com/item?id=49446422 得分：5 评论数：1
文章 URL: https://modelscope.cn/models/Qwen/Qwen3.8-Flash-Next 评论 URL: https://news.ycombinator.com/item?id=49432317 点数: 36 评论数: 9
文章 URL: https://www.thomsonreuters.com/en/press-releases/2026/august/thomson-reuters-leverages-its-world-class-data-assets-to-launch-its-own-frontier-model 评论 URL: https://news.ycombinator.com/item?id=49428318 点数: 10 评论数: 0
文章 URL： https://developers.openai.com/api/docs/models/gpt-5.6-sol 评论 URL： https://news.ycombinator.com/item?id=49396590 得分：10 评论数：5
文章 URL: https://nari-labs.com/blog/qwen3-tts-speed-cost-frontier/ 评论 URL: https://news.ycombinator.com/item?id=49389952 积分: 3 # 评论数: 1
Small language models are usually built like large ones and then squeezed onto a CPU afterwards. We did the opposite: we fixed the target first, one user, one token at a time, 4-bit weights, ordinary CPU, and chose the architecture to suit it. The result keeps full attention in only 6 of its 18 blocks. The other 12 use short convolutions whose memory is two timesteps wide no matter how long the conversation gets, so two thirds of the network never re-reads a growing cache. Trained from scratch on 59.9B tokens, the model scores 47.31 on a five-task benchmark against a bar of 42.20 that was fixed before training began. It beats GPT-2 124M, Pythia-160M, OPT-125M and GPT-neo-125M, all trained on three to six times more data, and exceeds MobileLLM-125M's published score despite that model seeing a trillion tokens. Validation bits-per-byte is 0.8685. To check the architecture rather than the training recipe, we trained a conventional all-attention model of the same size on the same data, and wrote down the winning condition before scoring either. The hybrid won the chosen quality metric by 0.81%, matched it on downstream tasks, produced a 6.3% smaller 4-bit file, and decoded 1.76x faster at 2048 tokens of context, 2.08x against an external model of similar size. In every measurement the speed advantage is near zero at an empty context and grows with length, which is what the mechanism predicts and what a merely leaner model would not show. A simple bandwidth calculation predicts only 1.17x, so memory volume alone does not explain the gap. We also report what did not work: an unmitigated 4-bit quality cost, roughly half the convolution channels ending up inert and impossible to remove, and a vocabulary larger than this model size warrants.
文章 URL： https://artificialanalysis.ai/models/glm-5-3 评论 URL： https://news.ycombinator.com/item?id=49353407 点数：32 评论数：7
OpenAI 正在加强对前沿 AI 模型的监控、对齐和安全性。了解新的保障措施如何引导模型开发的节奏。
文章 URL: https://openrouter.ai/openai/gpt-5.6-sol 评论 URL: https://news.ycombinator.com/item?id=49337602 积分: 31 评论数: 6
文章 URL： https://blog.roboflow.com/openai-gpt-5-6/ 评论 URL： https://news.ycombinator.com/item?id=49329575 积分：9 评论数：2
文章 URL： https://simonwillison.net/2026/Aug/16/qwen-38-27b/ 评论 URL： https://news.ycombinator.com/item?id=49324985 分数：17 评论数：3
文章 URL： https://koutian.is-a.dev/benchmark-radar/?view=leaderboard 评论 URL： https://news.ycombinator.com/item?id=49316791 点数：4 评论数：0
文章 URL： https://z.ai/blog/glm-5.3 评论 URL： https://news.ycombinator.com/item?id=49294997 点数：1084 评论数：526
文章 URL： https://www.mixedbread.com/blog/toast-1 评论 URL： https://news.ycombinator.com/item?id=49299746 积分：190 评论数：59
Gemini 2.5 Deep Think 在全球最负盛名的计算机编程竞赛中实现突破性表现，展现出抽象问题解决能力的深刻飞跃。
我们的新 Perch 模型帮助自然保护工作者更快地分析音频，以保护濒危物种，从夏威夷蜜旋木雀到珊瑚礁。
Genie 3 可以生成动态世界，你可以以每秒 24 帧的速度实时在其中导航，并在 720p 分辨率下保持数分钟的一致性。
推出首个用于为古代铭文提供语境化分析的模型，旨在帮助历史学家更好地解读、归属和复原残缺文本。
国际数学奥林匹克竞赛（“IMO”）是世界上最负盛名的青年数学家竞赛，自 1959 年以来每年举办一次。每个参赛国家由六名精英大学预科数学选手代表，他们竞相解答六道在代数、组合数学、几何和数论方面极其困难的问题。
新的 AI 模型整合了 PB 级地球观测数据，生成统一的数据表示，革新全球制图与监测
此前处于预览版的 Gemini 2.5 Flash-Lite 现已稳定并正式发布。这个高性价比模型以小巧体量提供高质量，并包含 2.5 系列功能，如 100 万 token 上下文窗口和多模态能力。
Gemma 3n 专为帮助塑造 Gemma 的开发者社区而设计。
我们宣布推出 MedGemma 系列中的全新多模态模型，这是我们用于健康 AI 开发的最强大开放模型。
介绍 T5Gemma，一个新的编码器-解码器 LLM 系列。
这款新的 AI 模型能够提供更高效、更准确且分辨率更高的全球天气预测。
AlphaFold 揭示了心脏病背后一种关键蛋白质的结构
探索 AlphaFold 如何加速科学发展，并推动全球生物学发现浪潮。
Gemini 3 Flash 以极低成本提供为速度而打造的前沿智能。
我们最新的 Veo 更新可生成生动、动态、自然且引人入胜的短片，并支持竖屏视频生成。
研究论文表明 Deep Think 在各领域的影响力日益增长
3.1 Pro 专为简单答案无法满足需求的任务而设计。
我们最新的图像生成模型具备先进的世界知识、生产就绪规格、主体一致性等能力，并全部以 Flash 速度实现。
Gemini 3.1 Flash-Lite 是我们迄今为止速度最快、成本效益最高的 Gemini 3 系列模型。
我们最新的语音模型提升了精确度并降低了延迟，使语音交互更加流畅、自然和精准。
Gemma 4：我们迄今最智能的开放模型，专为高级推理和智能体工作流打造。
Gemini Robotics ER 1.6：增强自主机器人的空间推理和多视角理解
我们最新的音频模型引入了细粒度音频标签，让你能够精确控制 AI 语音，以生成富有表现力的音频。
Gemini 3.5 旨在帮助你执行复杂的智能体式工作流。
了解我们的 WeatherNext AI 模型如何帮助预报员为社区争取前所未有的准备时间，以应对历史性的 Hurricane Melissa。
Google 推出 Gemini 3.5 Flash Cyber，这是一款轻量级网络安全模型，用于查找和修补漏洞。
我们正在推出新的 Gemini 模型，包括 Gemini 3.6 Flash、3.5 Flash-Lite 和 3.5 Flash Cyber。
Gemini Robotics ER 2 帮助机器人进行推理、协作并解决现实世界中的任务。它代表了机器人应用在视频理解、工具编排和多机器人协作方面的重大跃升。
This post describes four projects that share a common theme of enhancing or using generative models, a branch of unsupervised learning techniques in machine learning. In addition to describing our work, this post will tell you a bit more about generative models: what they are, why they are important, and where they might be going.
We’re open-sourcing OpenAI Baselines, our internal effort to reproduce reinforcement learning algorithms with performance on par with published results. We’ll release the algorithms over upcoming months; today’s release includes DQN and three of its variants.
We’re releasing a new class of reinforcement learning algorithms, Proximal Policy Optimization (PPO), which perform comparably or better than state-of-the-art approaches while being much simpler to implement and tune. PPO has become the default reinforcement learning algorithm at OpenAI because of its ease of use and good performance.
We’re releasing two new OpenAI Baselines implementations: ACKTR and A2C. A2C is a synchronous, deterministic variant of Asynchronous Advantage Actor Critic (A3C) which we’ve found gives equal performance. ACKTR is a more sample-efficient reinforcement learning algorithm than TRPO and A2C, and requires only slightly more computation than A2C per update.
Our team of five neural networks, OpenAI Five, has started to defeat amateur human teams at Dota 2.
The OpenAI Five Benchmark match is now over!
We’ve trained a human-like robot hand to manipulate physical objects with unprecedented dexterity.
Yesterday, OpenAI Five won a best-of-three against a team of 99.95th percentile Dota players: Blitz, Cap, Fogged, Merlini, and MoonMeander—four of whom have played Dota professionally—in front of a live audience and 100,000 concurrent livestream viewers.
OpenAI Five lost two games against top Dota 2 players at The International in Vancouver this week, maintaining a good chance of winning for the first 20–35 minutes of both games.
We’ve trained a large-scale unsupervised language model which generates coherent paragraphs of text, achieves state-of-the-art performance on many language modeling benchmarks, and performs rudimentary reading comprehension, machine translation, question answering, and summarization—all without task-specific training.
We’ll be holding our final live event for OpenAI Five at 11:30am PT on April 13.
OpenAI Five is the first AI to beat the world champions in an esports game, having won two back-to-back games versus the world champion Dota 2 team, OG, at Finals this weekend. Both OpenAI Five and DeepMind’s AlphaStar had previously beaten good pros privately but lost their live pro matches, making this also the first time an AI has beaten esports pros on livestream.
We’ve created MuseNet, a deep neural network that can generate 4-minute musical compositions with 10 different instruments, and can combine styles from country to Mozart to the Beatles. MuseNet was not explicitly programmed with our understanding of music, but instead discovered patterns of harmony, rhythm, and style by learning to predict the next token in hundreds of thousands of MIDI files. MuseNet uses the same general-purpose unsupervised technology as GPT-2, a large-scale transformer model trained to predict the next token in a sequence, whether audio or text.
We’re releasing the 774 million parameter GPT-2 language model after the release of our small 124M model in February, staged release of our medium 355M model in May, and subsequent research with partners and the AI community into the model’s potential for misuse and societal benefit. We’re also releasing an open-source legal agreement to make it easier for organizations to initiate model-sharing partnerships with each other, and are publishing a technical report about our experience in coordinating with the wider AI research community on publication norms.
We’ve trained a pair of neural networks to solve the Rubik’s Cube with a human-like robot hand. The neural networks are trained entirely in simulation, using the same reinforcement learning code as OpenAI Five paired with a new technique called Automatic Domain Randomization (ADR). The system can handle situations it never saw during training, such as being prodded by a stuffed giraffe. This shows that reinforcement learning isn’t just a tool for virtual tasks, but can solve physical-world problems requiring unprecedented dexterity.
As the final model release of GPT-2’s staged release, we’re releasing the largest version (1.5B parameters) of GPT-2 along with code and model weights to facilitate detection of outputs of GPT-2 models. While there have been larger language models released since August, we’ve continued with our original staged release plan in order to provide the community with a test case of a full staged release process. We hope that this test case will be useful to developers of future powerful models, and we’re actively continuing the conversation with the AI community on responsible publication.
我们推出了 Jukebox，这是一个能够以原始音频形式生成多种流派和艺术家风格音乐（包括基础歌声）的神经网络。我们将发布模型权重和代码，以及一个用于探索生成样本的工具。
OpenAI has agreed to license GPT-3 to Microsoft for their own products and services.
We’ve trained a neural network called DALL·E that creates images from text captions for a wide range of concepts expressible in natural language.
我们创建了 OpenAI Codex 的改进版本，这是我们的将自然语言转换为代码的 AI 系统，并将从今天起通过我们的 API 以私有测试版形式发布。
We’ve trained a system that solves grade school math problems with nearly twice the accuracy of a fine-tuned GPT-3 model. It solves about 90% as many problems as real kids: a small sample of 9-12 year olds scored 60% on a test from our dataset, while our system scored 55% on those same problems.
Fine-tune with a single command.
We’ve trained language models that are much better at following user intentions than GPT-3 while also making them more truthful and less toxic, using techniques developed through our alignment research. These InstructGPT models, which are trained with humans in the loop, are now deployed as the default language models on our API.
We’ve released new versions of GPT-3 and Codex which can edit or insert content into existing text, rather than just completing existing text.
Today, we are implementing a new technique so that DALL·E generates images of people that more accurately reflect the diversity of the world’s population.
We’ve trained and are open-sourcing a neural net called Whisper that approaches human level robustness and accuracy on English speech recognition.
We’ve trained a model called ChatGPT which interacts in a conversational way. The dialogue format makes it possible for ChatGPT to answer followup questions, admit its mistakes, challenge incorrect premises, and reject inappropriate requests.
我们很高兴宣布推出一款新的嵌入模型，它能力显著更强、更具成本效益，并且更易于使用。
我们创建了 GPT-4，这是 OpenAI 在扩展深度学习方面取得的最新里程碑。GPT-4 是一个大型多模态模型（接受图像和文本输入，输出文本），尽管在许多现实场景中能力不如人类，但在各种专业和学术基准测试中展现出人类水平的表现。
Developers can now bring their own data to customize GPT-3.5 Turbo for their use cases.
我们宣布公开招募 OpenAI Red Teaming Network，并邀请有兴趣提升 OpenAI 模型安全性的领域专家加入我们的工作。
我们正在分享 Voice Engine 小规模预览版中的经验教训，这是一个用于创建自定义语音的模型。
Harvey 与 OpenAI 合作，为法律专业人士构建定制训练模型。
我们很高兴宣布在亚洲设立首个办公室，并发布一款针对日语优化的 GPT-4 自定义模型。
我们宣布推出 GPT-4 Omni，这是我们的全新旗舰模型，能够跨音频、视觉和文本进行实时推理。
探索我们的文本转语音模型背后的技术。
CriticGPT 是一个基于 GPT-4 的模型，它会撰写对 ChatGPT 回复的评析，帮助人类训练员在 RLHF 过程中发现错误
Economist Tyler Cowen explains how OpenAI o1 tackles complex economic questions.
Scott Wu, CEO and Co-Founder of Cognition, explains how OpenAI o1 makes coding decisions in a more human-like way.
OpenAI o1 Contributions
Advancing cost-efficient reasoning
Building smarter maps with GPT-4o vision fine-tuning
本报告概述了在发布 OpenAI o1 和 o1-mini 之前开展的安全工作，包括根据我们的 Preparedness Framework 进行的外部红队测试和前沿风险评估。
Sora 是 OpenAI 的视频生成模型，旨在接收文本、图像和视频输入，并生成新的视频作为输出。Sora 基于 DALL-E 和 GPT 模型的经验构建，旨在为人们提供更丰富的工具，用于讲故事和创意表达。
OpenAI 最新的推理模型系列将被美国顶尖科学家用于推动科学突破。
本报告概述了针对 OpenAI o3-mini 模型开展的安全工作，包括安全评估、外部红队测试以及 Preparedness Framework 评估。
我们正在发布 OpenAI GPT‑4.5 的研究预览版，这是我们迄今为止规模最大、知识最丰富的模型。
4o image generation 是一种新的图像生成方法，相比我们早期的 DALL·E 3 系列模型能力显著更强。它可以创建逼真的输出。它可以将图像作为输入并对其进行转换。
在 API 中推出 GPT-4.1——一个全新的模型家族，带来全面改进，包括在编码、指令遵循和长上下文理解方面的重大提升。我们还发布了首个 nano 模型。即日起面向全球开发者开放。
OpenAI o3 和 OpenAI o4-mini 将最先进的推理能力与完整的工具能力相结合——网页浏览、Python、图像和文件分析、图像生成、画布、自动化、文件搜索和记忆。
迄今为止我们最智能、能力最强的模型，并具备完整的工具访问权限
我们最新的图像生成模型现已通过 ‘gpt-image-1’ 在 API 中提供，使开发者和企业能够将专业级、可定制的视觉内容直接构建到自己的工具和平台中。
我们已回滚 ChatGPT 中上周的 GPT‑4o 更新，因此人们现在使用的是一个行为更均衡的早期版本。我们移除的这次更新过于奉承或迎合——通常被形容为谄媚。
深入探讨我们的发现、哪里出了问题，以及我们未来将做出的改变。
AI 的下一个前沿不仅关乎能力，也关乎谁能使用它。我们致力于让尽可能多的人掌握 AI，这正是驱动我们的使命。今天发布我们迄今最强大的开放权重模型，是向前迈出的重要一步，使先进 AI 在全球范围内更加开放、灵活且触手可及。
我们介绍 gpt-oss-120b 和 gpt-oss-20b，这两个开放权重推理模型根据 Apache 2.0 许可证和我们的 gpt-oss 使用政策提供。
我们正在发布 gpt-oss-120b 和 gpt-oss-20b——两个先进的开放权重语言模型，能够以低成本提供强大的实际性能。这些模型采用灵活的 Apache 2.0 许可证，在推理任务上的表现优于同等规模的开放模型，展现出强大的工具使用能力，并针对在消费级硬件上的高效部署进行了优化。
我们正在推出 GPT‑5，这是我们迄今为止最出色的 AI 系统。GPT‑5 在智能方面相比我们以往所有模型都有显著飞跃，在编码、数学、写作、健康、视觉感知等领域具备最先进的性能。
看看一群顶尖开发者首次使用 GPT-5 的体验。
这份 GPT-5 系统卡说明了统一的模型路由系统如何通过 gpt-5-main、gpt-5-thinking 以及 gpt-5-thinking-nano 等轻量版本，为不同任务和开发者用途优化，从而实现快速且智能的响应。
了解 OpenAI 在 GPT-5 中新的 safe-completions 方法如何提升 AI 回复的安全性和有用性——超越硬性拒绝，转向细致入微、以输出为中心的安全训练，以处理双重用途提示。
了解 GPT-5 如何用于医学研究。
了解 GPT-5 如何在编码和设计中开启新的可能性。
GPT-5 是 OpenAI 最先进的模型，正在智能工作新时代变革企业 AI、自动化和员工生产力。
在我们的 API 平台中推出 GPT-5——提供高水平推理性能、面向开发者的新控制能力，以及在真实编码任务上的一流表现。
了解专用 AI 模型 GPT-4b micro 如何帮助 OpenAI 和 Retro Bio 设计出用于干细胞疗法和长寿研究的更有效蛋白质。
我们正在发布一款更先进的语音到语音模型，以及新的 API 功能，包括 MCP 服务器支持、图像输入和 SIP 电话呼叫支持。
这份 GPT-5 系统卡补遗介绍了一个新模型：GPT-5-Codex，这是 GPT-5 的一个版本，针对 Codex 中的智能体式编码进行了进一步优化。GPT-5-Codex 会根据任务复杂度更动态地调整其思考力度，对简单的对话查询或小任务快速响应，同时在更复杂的任务上能够独立工作更长时间。
我们最新的视频生成模型相比以往系统，在物理准确性、真实感和可控性方面都有提升。它还具备同步对白和音效功能。可在新的 Sora 应用中使用它进行创作。
Sora 2 是我们全新的最先进视频和音频生成模型。在 Sora 的基础上，这一新模型引入了以往视频模型难以实现的能力，例如更准确的物理效果、更清晰的真实感、同步音频、增强的可控性，以及更广泛的风格范围。
本系统卡详细介绍了 GPT-5 在处理敏感对话方面的改进，包括针对情感依赖、心理健康和越狱抵抗能力的新基准。
OpenAI 推出 gpt-oss-safeguard——用于安全分类的开放权重推理模型，使开发者能够应用并迭代自定义策略。
我们正在升级 GPT-5 系列，推出更亲和、更强大的模型，以及自定义 ChatGPT 语气和风格的新方式。GPT-5.1 将于今天开始向付费用户逐步推出。
本 GPT-5 系统卡附录提供了 GPT-5.1 Instant 和 Thinking 的最新安全指标，包括针对心理健康和情感依赖的新评估。
GPT-5.1 现已在 API 中可用，带来更快的自适应推理、扩展的提示缓存、改进的编码性能，以及新的 apply_patch 和 shell 工具。
推出 GPT-5.1-Codex-Max，这是面向 Codex 的更快、更智能的智能体编程模型。该模型专为长时间运行、项目规模的工作而设计，具备增强的推理能力和 token 效率。
本系统卡概述了为 GPT‑5.1-CodexMax 实施的全面安全措施。它详细介绍了模型层面的缓解措施，例如针对有害任务和提示注入的专门安全训练，以及产品层面的缓解措施，例如智能体沙箱和可配置的网络访问。
OpenAI introduces the first research cases showing how GPT-5 accelerates scientific progress across math, physics, biology, and computer science. Explore how AI and researchers collaborate to generate proofs, uncover new insights, and reshape the pace of discovery.
UCLA 教授 Ernest Ryu 与 GPT-5 解决了优化理论中的一个关键问题，展示了 AI 在加速数学发现中的作用。
GPT-5.2 是我们面向日常专业工作的最先进前沿模型，具备最先进的推理、长上下文理解、编码和视觉能力。可在 ChatGPT 和 OpenAI API 中使用它，以支持更快速、更可靠的智能体工作流。
GPT-5.2 是 GPT-5 系列中的最新模型家族。针对这些模型的综合安全缓解方法与 GPT-5 System Card 和 GPT-5.1 System Card 中描述的方法大体相同。与 OpenAI 的其他模型一样，GPT-5.2 模型基于多样化数据集进行训练，包括互联网上公开可用的信息、我们与第三方合作获取的信息，以及我们的用户或人类训练师和研究人员提供或生成的信息。
GPT-5.2 是 OpenAI 迄今为止在数学和科学方面最强的模型，在 GPQA Diamond 和 FrontierMath 等基准测试中取得了新的最先进结果。本文展示了这些提升如何转化为真实的研究进展，包括解决一个开放的理论问题以及生成可靠的数学证明。
GPT-5.2-Codex 是 OpenAI 最先进的编码模型，具备长程推理、大规模代码转换和增强的网络安全能力。
OpenAI 正在更新其 Model Spec，加入新的未满 18 岁原则，定义 ChatGPT 应如何基于发展科学，为青少年提供安全且适龄的指导。此次更新强化了防护措施，明确了模型在较高风险情境中的预期行为，并建立在我们更广泛的工作之上，以提升 ChatGPT 中的青少年安全。
GPT‑5.3-Codex 是迄今为止能力最强的智能体式编码模型，结合了 GPT‑5.2-Codex 的前沿编码性能与 GPT‑5.2 的推理和专业知识能力。
GPT-5.3-Codex 是一个 Codex 原生智能体，将前沿编码性能与通用推理相结合，以支持长期的、现实世界中的技术工作。
推出 GPT-5.3-Codex-Spark——我们的首个实时编程模型。生成速度快 15 倍，支持 128k 上下文，现已面向 ChatGPT Pro 用户推出研究预览版。
一篇新的预印本显示，GPT-5.2 提出了一个新的胶子振幅公式，随后由 OpenAI 和学术合作者进行了形式化证明和验证。
我们分享我们的 AI 模型针对 First Proof 数学挑战的证明尝试，以测试在专家级问题上的研究级推理能力。
推出 GPT-5.4，这是 OpenAI 面向专业工作的最强大且最高效的前沿模型，具备最先进的编码、计算机使用、工具搜索能力，以及 100 万 token 上下文。
GPT-5.4 mini 和 nano 是 GPT-5.4 的更小、更快版本，针对编码、工具使用、多模态推理以及高容量 API 和子代理工作负载进行了优化。
随着 AI 网络安全能力不断提升，OpenAI 扩展了其 Trusted Access for Cyber 计划，向经过审核的防御者推出 GPT-5.4-Cyber，并加强安全防护措施。
OpenAI 推出 GPT-Rosalind，这是一款前沿推理模型，旨在加速药物发现、基因组学分析、蛋白质推理和科学研究工作流程。
OpenAI Privacy Filter 是一个开放权重模型，用于以最先进的准确率检测和编辑文本中的个人身份信息（PII）。
推出 GPT-5.5，这是我们迄今最智能的模型——速度更快、能力更强，专为跨工具的编码、研究和数据分析等复杂任务而打造。
哥布林输出如何在 AI 模型中传播：GPT-5 行为中由个性驱动的怪癖背后的时间线、根本原因和修复措施。
GPT-5.5 Instant 更新了 ChatGPT 的默认模型，带来更智能、更准确的回答，减少幻觉，并改进个性化控制。
探索 OpenAI API 中新的实时语音模型，它们能够推理、翻译和转录语音，从而实现更自然、更智能的语音体验。
一个 OpenAI 模型解决了已有 80 年历史的单位距离问题，推翻了离散几何中的一个重要猜想，标志着 AI 驱动数学的一个里程碑。
GPT-Rosalind 通过增强的生物学推理、药物化学专业知识、基因组学分析和实验工作流程能力，推动生命科学研究。
GPT-5 Pro 帮助解开了一个历时 3 年的免疫学谜题，为 T 细胞行为提供了见解。这一突破可能有助于癌症和自身免疫研究。
OpenAI 预览 GPT-5.6 Sol，这是一款在编码、科学和网络安全方面具备更强能力的下一代模型，并配备其最先进的安全技术栈。
新一代语音模型，专为自然的人类与 AI 交互而打造，现已为 ChatGPT Voice 提供支持。
每个 token 带来更多智能，每美元获得更强性能，并可按需提供更多能力，助力你最艰巨的工作。
GPT-5.6 提升了模型、推理和智能体工作流中的 AI 效率，帮助以每一美元提供更有用的智能。
两项 API 设置如何通过保留推理并启用压缩来提升 GPT-5.6 在 ARC-AGI-3 上的性能，从而提高分数和效率。
了解 Luna 和 Terra 更低的 GPT‑5.6 定价，以及 OpenAI 更高效的模型如何帮助企业大规模部署 AI 工作流。
OpenAI 解释了近期第三方网络安全评估事件，并概述了旨在加强 AI 模型测试与评估的新防护措施。
认识 GPT-5.6-Cyber，这是 OpenAI 面向网络安全的专用模型，可通过 Daybreak Red 使用，用于授权的漏洞研究、漏洞利用验证和安全测试。
了解初创公司如何使用 GPT-5.6，通过更智能的模型选择和新的 Responses API 功能，更快速、更具成本效益地构建 AI 智能体。