webAI Releases TwIL-8B, an 8.8B Reasoning Model Scoring 98.2 on MATH-500

webAI released TwIL-8B, an 8.8B reasoning model post-trained from IBM Granite 4.2 8B. It posts 98.2 on MATH-500 and 81.6 on LiveCodeBench, but trails similar-size models on knowledge benchmarks. The post webAI Releases TwIL-8B, an 8.8B Reasoning Model Scoring 98.2 on MATH-500 appeared first on MarkTechPost .