Event ·Multi-source

webAI开源3B推理模型TwiL-LM3，4项基准超GPT-OSS-120B

webAI Intelligence Lab 开源了首个形式化推理模型 TwiL-LM3，参数量仅 3B。在 5 项形式化推理基准中，TwiL-LM3 有 4 项超过 OpenAI 的 GPT-OSS-120B，参数量约为后者的 1/40，推理速度快 2.6 倍。该模型基于 webAI 自有的验证数据集训练，而非抓取互联网数据，可在树莓派或 iPhone 等消费级硬件上运行。

Current read

3B 的小模型在推理上干翻了 120B 的大家伙，还能跑在树莓派上，开源了直接就能玩。

11 sources58° AI heatLast updated 2026/8/11 02:42:41

Evidence

11 sources

Primary sourceshao__meng

2026/8/11 02:42:41View source

Related source 1Decoder

2026/8/11 15:07:44View source

Related source 2Thomas Wolf

2026/8/9 09:28:04View source

Related source 3arXiv: OpenAI

2026/8/9 19:45:35View source

Related source 4向阳乔木

2026/8/10 14:29:24View source

Related source 5IT之家

2026/8/10 14:59:43View source

Related source 6OpenAI

2026/8/10 17:16:53View source

Related source 7Sam Altman

2026/8/10 18:25:03View source

Related source 8The Rundown AI

2026/8/11 10:30:24View source

Related source 9Amjad Masad

2026/8/9 15:05:10View source

Related source 10Greg Brockman

2026/8/10 17:27:11View source

Conflict check

Dedup data only shows that these sources discuss the same event, not that they agree. There is no structured supporting or refuting evidence yet, so no conflict is inferred automatically.

Open the story