Event ·Multi-source

OpenAI 用 GPT-5.6 Sol 优化自身推理，服务成本降20%

OpenAI 于7月29日部署 GPT-5.6 Sol 模型优化 GPT-5.6 系列推理链路。该模型通过 Codex 分析生产流量并学习 Triton 和 Gluon 语言，自主改写 GPU 内核，使端到端服务成本降低最多20%。此外，优化推测性解码后，token 生成效率提升超过15%。OpenAI 还使用开源工具 FpSan 验证内核正确性。

Current read

OpenAI 让自家模型 GPT-5.6 Sol 反过来优化自己的推理，成本降了20%，token 生成快了15%，挺有意思的。

11 sources83° AI heatLast updated 2026/7/30 06:06:15

Evidence

11 sources

Primary sourceIT之家

2026/7/30 06:06:15View source

Related source 1Greg Brockman

2026/7/29 21:31:16View source

Related source 2@OpenAIDevs

2026/7/29 21:33:09View source

Related source 3宝玉

2026/7/28 22:10:39View source

Related source 4歸藏(guizang.ai)

2026/7/29 04:11:22View source

Related source 5OpenAI

2026/7/29 21:21:29View source

Related source 6Simon Willison

2026/7/30 01:34:49View source

Related source 7Sam Altman

2026/7/30 17:29:25View source

Related source 8小互

2026/8/1 02:46:34View source

Related source 9Amjad Masad

2026/7/28 07:09:33View source

Related source 10arXiv: OpenAI

2026/7/28 15:33:54View 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