Event ·Multi-source

Cognition 发布 SWE-2 模型，在代码生成任务上表现接近 Fable 5.1

Cognition 公司的 SWE-2 模型在 FrontierCode 1.1 Main1 基准测试中取得了 50.0% 的成绩，仅比 Fable 5.1 低 1 分，同时成本降低了 64%。该模型通过扩展强化学习至数千亿参数，在能力和成本之间取得了新的平衡。

Current read

Cognition 推出了 SWE-2，这个模型在代码生成上很厉害，和 Fable 5.1 差不多，但便宜很多。

5 sources58° AI heatLast updated 2026/9/10 16:00:30

Evidence

5 sources

Primary sourceelvis

2026/9/10 16:00:30View source

Related source 1Cognition

2026/9/10 15:21:01View source

Related source 2The Rundown AI

2026/9/10 16:44:04View source

Related source 3Windsurf

2026/9/10 16:52:47View source

Related source 4Fireworks AI

2026/9/10 22:40:06View 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