● ResearchNov 7, 2026 · 15:227 sources
Hugging Face Finds webAI's TwIL-8B Model Beats LFM3-1.7B in AI Agent Tests
From marketbrief.now
AI;DR
Hugging Face researchers found that webAI's TwIL-8B model outperformed Liquid AI's LFM3-1.7B across three agent harnesses during preliminary baseline evaluations. TwIL-8B is an 8.8B parameter open-weight model built on IBM's Granite 4.2 8B that was designed to run on consumer CPUs and graphics cards.
Read originalSign in to suggest
Sources
+ 4 more
- SUPPORT@rohanpaul_ai“TwIL-8B fits in a 4.98 GiB quantized file that runs on a CPU or modest graphics card”x.com
- SUPPORT@yb2698“the Hugging Face research team is already pushing its capabilities further”x.com
- SUPPORT@hesamation“this model from webAI is small enough to run locally… and make a lot of expensive API calls unnecessary”x.com
- SUPPORT@omarsar0“smaller models will play a significant role in more complex agentic workflows like scientific discovery and software factories”x.com