OpenAI released GPT-6.1 Sol at DevDay on September 29. It says the model nearly matches GPT-6 Astra on agentic coding at a fifth of Astra's token price. The API price stays at GPT-6 Sol's $2 per million input tokens and $10 per million output tokens. We graded it the next day as the successor to GPT-6 Sol.
What the results show
- It wins back what GPT-6 Sol lost. GPT-6.1 Sol moves from fifth to third on the board. It writes better plans than GPT-6 Sol and does better on one-line bug reports. It is a bit lighter on the plan too, at about 6% of a ChatGPT Plus week for a run of our suites.
- Claude Opus 5.5 still does more. Claude Opus 5.5 grades higher on real issues and on one-line bug reports. Sol is lighter on the plan though. At the same monthly price a run takes about 40% of Sol's five-hour limit against about 90% for Opus. Over the week it is about 6% against 10%.
Our take: if GPT-6.1 Sol had come out two weeks ago, it would have been a great release. On that week's board it would have sat level with Claude Fable 5.1 at the top. But Opus 5.5 and Sonnet 5.5 arrived in between. Opus is the better model for hard work and Sonnet is the better one for the small stuff. Together they beat the Sol and Luna pair at the same price. The OpenAI pair still goes further on the plan. If quota is what runs out first for you, that counts for something. For most coding work we would pick the Claude pair.
Grades are comparable within the same suite version. See the full GPT-6.1 Sol scorecard or follow the family over time on Trends.