Event ·Multi-source

OpenAI 推出 GPT-6 Luna:低成本档在 DeepSWE v1.1 拿 66.6%

OpenAI 发布 GPT-6 Luna,定位为快速、低成本档位。官方报告其在 DeepSWE v1.1 基准上得分 66.6%,与 Claude Opus 5 和 Fable 5 在 medium effort 档的成绩相当,但每个任务的成本降低 93-96%。在更高 effort 档下,Luna 的事实性(factuality)可匹敌 GPT-5.6 Sol,成本约为后者的 1/100。该模型已可通过 OpenRouter 调用。

Current read

OpenAI 出了个便宜档的 GPT-6 Luna,跑分追平 Claude Opus 5,但只要零头的价格,看重成本的可以去 OpenRouter 试试。

11 sources83° AI heatLast updated 2026/9/22 18:23:41

Evidence

11 sources
Related source 3shao__meng
View source
Related source 6Simon Willison’s Weblog
View source
Related source 7lmarena.ai
View source
Related source 8SuperTechFans
View 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.