GPT-6.1 Sol approaches Astra on selected benchmarks, with API input pricing from $2 per million tokens, but it is not yet available in Chat.
OpenAI has released GPT-6.1 Sol, an upgrade aimed at agentic coding, computer use and professional work at a lower cost than GPT-6 Astra. Developers can access it through the API under the identifier gpt-6.1-sol. It is also available to Plus, Pro, Business, Enterprise and Edu users in ChatGPT Work and Codex.
Standard API pricing is $2 per million input tokens, $0.10 per million cached input tokens and $10 per million output tokens. OpenAI says the standard input and output rates are one-fifth of Astra’s, positioning Sol for workloads where an agent may repeatedly read context, invoke tools and produce substantial outputs.
In OpenAI’s evaluations, GPT-6.1 Sol matched Astra on DeepSWE v1.1 and exceeded GPT-6 Sol’s best result by 6.4 percentage points. At medium reasoning effort, it improved on GPT-6 Sol by 4.8 points in AutomationBench. On the offline OSWorld 2.0 set, it gained seven points at maximum effort and finished 2.1 points behind Astra.
Astra retained a clear advantage on the most difficult scientific research tasks. GPT-6.1 Sol more than doubled its predecessor’s Terminal-Bench Science 0.1 score while averaging $5.47 per task, but Astra remained the strongest model tested, with a score of 68.1%. OpenAI therefore does not present the less expensive model as an across-the-board replacement for Astra.
OpenAI also reported a reduction in factual errors at low reasoning effort: the share of responses containing at least one error fell from 11.4% to 7.7%. That result should not be treated as a measure of ordinary conversations because the evaluation used deliberately difficult, de-identified exchanges in which users had previously flagged errors. The company likewise said its safety evaluations focused on challenging situations rather than typical failure rates.
At publication, GPT-6.1 Sol was not yet available in the regular Chat experience. OpenAI also warns that research and API evaluation results may differ from production ChatGPT behavior because system prompts, available tools, reasoning effort and other conditions can vary.
For teams already running coding agents or interface automation with GPT-6 Sol, the release offers a potential quality improvement without moving to Astra’s higher token rates. The practical choice still requires testing on each workflow: the published performance gap varies by task, while the cost of an agent run depends on context length, output volume and how many iterations it performs.
| Token type | Price per 1 million tokens |
|---|---|
| Input | $2 |
| Cached input | $0.10 |
| Output | $10 |
Sources
Originally published by AI Feed: OpenAI Releases GPT-6.1 Sol for Agentic Coding and Computer Use.
Top comments (0)