ReleasesOct 1, 2026 · 08:232 sources
webAI's 3.66B TwIL-LM3-Pro Beats VibeThinker by 35% in Formal Logic
From marketbrief.now
Austin-based webAI released the TwIL-LM3-Pro, a 3.66 billion parameter reasoning model that scored 35% higher than VibeThinker-3B, 24% higher than Qwen3.5-4B, and 47% higher than Liquid AI's LFM2.5-8B-A1B in formal logic, while nearly matching gpt-oss-120b on GSM8K with 94.3% accuracy. The model is distributed as a 2.09GiB Q4 build running via llama.cpp on CPUs or local GPUs, and webAI is developing Meridian, a series of frontier-class on-device models.
Read the sourceSign in to suggest
Sources
One ranked edition, every morning
The stories that matter today. Free, cancel anytime.
Telegram @aidr_today