OpenAI: GPT-5.6 Sol autonomously fine-tunes smaller model with vague prompt

OpenAI's GPT-5.6 Sol autonomously fine-tuned the smaller Luna model using an underspecified prompt.

Published 3j1 sourceNotable

≈ 17sSpeed

The fact

In OpenAI's internal RSI benchmark, Sol outperforms GPT-5.5 by 16.2 points.

OpenAI believes an 'automated researcher' is now within reach.

Explore this topic

Cross-sources — 1 source

Auto-synthesis from 1 media source · identified on July 10, 2026