Event ·Multi-source

OpenAI 发布 GPT-5.6 系列模型，性能大幅提升且成本降低

GPT-5.6 Sol 在 Agents' Last Exam 上得分 53.6，领先 Claude Fable 5 达 13.1 分。中等推理模式下仅用约四分之一成本即领先 11.4 分。Terra 和 Luna 以约十六分之一成本超越 Fable 5。在 Artificial Analysis Coding Agent Index 上 Sol 取得 80 分，领先 Fable 5 2.8 分，token 和成本大幅降低。新推出 ultra 模式默认协调四个智能体并行工作，提升浏览和证券基准表现。安全方面经过广泛红队测试和自动化评估。

Current read

OpenAI 刚发了 GPT-5.6，三个型号性能都爆表，尤其 Sol 在编码和推理测试里吊打 Claude Fable 5，成本还低很多，编程和复杂任务都更强了。

11 sources92° AI heatLast updated 2026/7/11 00:01:15

Evidence

11 sources

Primary sourceSuperTechFans

2026/7/11 00:01:15View source

Related source 1Simon Willison’s Weblog

2026/7/9 19:46:38View source

Related source 2marktechpost

2026/7/9 20:45:42View source

Related source 3OpenAI

2026/7/9 17:30:39View source

Related source 4The Rundown AI

2026/7/9 17:38:43View source

Related source 5AI Will

2026/7/10 04:21:54View source

Related source 6IT之家

2026/7/9 22:46:08View source

Related source 7@OpenAIDevs

2026/7/9 17:41:15View source

Related source 8Decoder

2026/7/9 18:37:51View source

Related source 9Poe

2026/7/9 20:18:18View source

Related source 10shao__meng

2026/7/10 09:02:01View 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