Event ·Multi-source

OpenAI 称 GPT-5.6 Sol 可自动化后训练 Luna 模型

GPT-5.6 Sol 能充当“自动化研究员”，为较小模型 Luna 定制完整后训练方案。在内部评估套件“聚合 RSI”上，GPT-5.6 Sol 比 GPT-5.5 高出 16.2 个百分点。活跃研究员人均日 token 产出较 GPT-5.5 翻倍以上，单个研究员 pull request 和实验数量均上升。

Current read

OpenAI 新出的 GPT-5.6 Sol 能自己训练小模型Luna，研究效率比上一代高16%，研究员一天干的活翻倍了。

11 sources73° AI heatLast updated 2026/7/11 07:14:57

Evidence

11 sources

Primary sourceIT之家

2026/7/11 07:14:57View source

Related source 1OpenAI

2026/7/9 10:45:06View source

Related source 2Greg Brockman

2026/7/9 17:26:59View source

Related source 3Simon Willison’s Weblog

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

Related source 4marktechpost

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

Related source 5LlamaIndex

2026/7/9 22:48:21View source

Related source 6AI Will

2026/7/10 04:21:47View source

Related source 7Decoder

2026/7/10 18:43:47View source

Related source 8lmarena.ai

2026/7/10 20:04:20View source

Related source 9SuperTechFans

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

Related source 10elvis

2026/7/9 17:25:27View 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