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

MarkTechPostNovember 4, 2026en

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 .

This is a short summary published by AI Global Wire. The full article is owned and hosted by MarkTechPost — open it there to read it in full.

• Verktyg