webAI 发布 3.66B 参数形式逻辑模型 TwIL-LM3-Pro，可在笔记本本地运行。其测试中，该模型在形式逻辑综合基准上与 Qwen3-8B 大致持平，参数量不到后者一半，并在全部六项形式任务上领先公开的 VibeThinker-3B 检查点。模型基于 IBM Granite 4.2 后训练，Q4 GGUF 权重仅 2.09 GiB，可通过 llama.cpp 本地推理。
webAI has released a 3.66B model for formal logic that can run locally on a laptop.
In webAI’s tests, TwIL-LM3-Pro is roughly level with Qwen3-8B on the headline formal-logic composite, with less than half the parameters. It also leads the public VibeThinker-3B checkpoint on all six reported formal tasks.
The team specialized IBM’s Granite 4.2 through post-training to check whether conclusions follow from premises, translate statements into logic and critique formal proofs.
The Q4 GGUF weight file is just 2.09 GiB and can run locally via llama.cpp. You can keep inference on your own machine.
Weights and model card (non-commercial license):
https://huggingface.co/webAI-Official/TwIL-LM3-Pro
@thewebai @Davidstout
来源：Chubby♨️ · x.com