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, scoring 98.2 on MATH-500 and 81.6 on LiveCodeBench, though it trails similar-size models on knowledge benchmarks.

Read the full story at MarkTechPost ↗

Timeline · 1 report

2026-11-04 22:25 · MarkTechPost
webAI Releases TwIL-8B, an 8.8B Reasoning Model Scoring 98.2 on MATH-500

More stories

Critical Langflow Flaws Allow Unauthenticated Code Execution, IBM Warns — The Hacker News (security)
AI uncovers major web security flaw as new voice model emerges — Mixture of Experts (IBM)
IBM podcast examines self-replicating AI, medical chatbots and AI agents — Mixture of Experts (IBM)
Anthropic's FlashProve prototype verifies which AI model generated a response without exposing weights — Runway Research
OpenAI to Acquire Developer Tools Startup Pierre Computer — OpenAI News
Democratic Midterm Wins Cloud Outlook for AI Data Center Expansion — DigiTimes Asia
Musk says Grok Bot costs are 'unsustainably high' as usage limits tighten — Forbes AI
Exit Poll Finds AI and Data Centers Weighed Heavily in Midterm Votes — Barron's

Get the daily brief of stories like this at 6:30 every morning →