According to OpenAI, GPT-6.1 Sol, an upgrade to its mid-tier GPT-6 Sol model, now approaches the intelligence of the flagship GPT-6 Astra on coding, computer-use and professional-document tasks at roughly a fifth of Astra's standard token price. It is live today in ChatGPT Work and Codex for Plus, Pro, Business, Enterprise and Edu plans, and in the API as gpt-6.1-sol at $2 per million input tokens, $10 per million output tokens, and $0.10 per million cached input tokens, a price OpenAI says is 95% below its standard input rate.
The benchmark claims
On DeepSWE v1.1, a test of long-horizon software-engineering tasks, OpenAI says GPT-6.1 Sol matches Astra's score at about a fifth of the cost, and beats GPT-6 Sol's best result by 6.4 percentage points while using less reasoning effort. On OSWorld 2.0's offline computer-use benchmark, OpenAI says GPT-6.1 Sol comes within 2.1 points of Astra at maximum reasoning effort, at roughly a seventh of the cost per task. On AutomationBench, a 47-tool business-workflow test, OpenAI says GPT-6.1 Sol scores 2.2 points above Claude Opus 5.5 at medium effort and about a third of the cost, though it notes the Opus 5.5 figure omits the cost of fallbacks to other models, which it says happened on roughly 40% of tasks.
OpenAI also reports a factuality gain: on prompts drawn from conversations where users had flagged an earlier model's error, GPT-6.1 Sol's error rate at low reasoning effort falls from 11.4% for GPT-6 Sol to 7.7%, landing within 1.9 points of Astra's rate.
Caveats and what has not shipped
Every number above is OpenAI's own, run in its research environment or via its API rather than production ChatGPT, and the competitor scores are pulled from publicly available reports rather than a shared test. Astra still holds the top score OpenAI tested on the hardest task, a scientific-research benchmark, at 68.1%, and the company says Astra remains the right choice for the most demanding work. GPT-6.1 Sol is not yet available in ChatGPT's standard Chat mode, and a faster "Ultrafast" variant, which OpenAI says will run up to eight times faster in Codex, has not shipped yet either.