Event ·Multi-source

Opus 5 在 FrontierCode 1.1 上得分63.6%，接近Fable 5但成本减半

FrontierCode 1.1 是 Cognition 发布的评估真实工程任务的基准，考核代码合并性和质量。Opus 5 在该基准上取得 63.6% 的分数，扩展测试通过率 69.6%，性能接近 Fable 5 而成本仅为后者的一半。在 Devin 环境中，Opus 5 在困难调试和根因分析任务上表现突出。

Current read

Opus 5 在代码工程基准上接近顶级模型，但价格只要一半，做调试分析特别强。

11 sources58° AI heatLast updated 2026/7/24 17:07:20

Evidence

11 sources

Primary sourceCognition

2026/7/24 17:07:20View source

Related source 1Alex Albert

2026/7/24 17:14:15View source

Related source 2orange.ai

2026/7/24 23:44:16View source

Related source 3Claude

2026/7/24 16:59:54View source

Related source 4techcrunch

2026/7/24 17:00:00View source

Related source 5Decoder

2026/7/26 09:43:02View source

Related source 6歸藏(guizang.ai)

2026/7/23 12:22:43View source

Related source 7Jerry Liu

2026/7/24 22:52:17View source

Related source 8Nous Research

2026/7/24 23:19:42View source

Related source 9Simon Willison’s Weblog

2026/7/25 00:42:59View source

Related source 10OpenRouter

2026/7/25 17:23:46View 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