OpenAI released GPT-6.1 Sol on September 29, 2026, an upgrade to the GPT-6 Sol model it shipped a week earlier. The pitch: performance that nearly matches GPT-6 Astra on agentic coding, computer use, and professional work, at one-fifth of Astra's standard input and output prices.
Sol's price did not change. It is $2 per million input tokens and $10 per million output tokens, the same as GPT-6 Sol. Astra is listed at $10 input and $50 output. Cached input is $0.10 per million tokens, 95% below the standard input rate.
Where you can use it
The API model ID is gpt-6.1-sol. OpenAI says it is available to Plus, Pro, Business, Enterprise, and Edu users in ChatGPT Work and Codex. It is not yet available in regular Chat, the same rollout pattern as the earlier GPT-6 Sol and Luna release. DataCamp also lists availability through OpenRouter, Vercel AI Gateway, and GitHub Copilot Pro+, and a 1.05M-token context window with up to 128K output tokens. Requests above 272K input tokens are billed at a higher long-context rate.
OpenAI's benchmark claims
OpenAI frames the release around cost per task, not just scores:
| Benchmark | OpenAI-reported result for GPT-6.1 Sol |
|---|---|
| DeepSWE v1.1 | Matches GPT-6 Astra at one-fifth the cost; 6.4 points above GPT-6 Sol |
| GDP.pdf | Approaches Astra and beats Claude Opus 5.5 at half the cost per task |
| AutomationBench | 4.8 points above GPT-6 Sol and 2.2 above Opus 5.5, at one-third the cost |
| OSWorld 2.0 | Within 2.1 points of Astra at max effort, 7 points above GPT-6 Sol, at one-seventh the cost |
| Terminal-Bench Science 0.1 | Doubles GPT-6 Sol's score at $5.47 per task versus roughly $23 for the comparison models |
| Factuality | Error rate of 7.7% at low effort, down from 11.4%, within 1.9 points of Astra |
These are vendor-reported figures, and several depend on OpenAI's chosen effort settings and cost estimates. The relevant comparison for most teams is not Astra on a leaderboard but whether Sol completes your own tasks at an acceptable rate for the money.
The Ultrafast tier
OpenAI also announced an Ultrafast option with up to 8x faster token generation in Codex. VentureBeat reports it reaches about 300 tokens per second through the API and is priced at roughly six times standard rates, which would put GPT-6.1 Sol Ultrafast near $12 input and $60 output per million tokens (VentureBeat's derived figure, not an OpenAI price sheet). GPT-6 Astra Ultrafast is available now for Pro 500 and Enterprise; the Sol version is listed as coming soon.
Safety notes
OpenAI's system card addendum says GPT-6.1 Sol beat GPT-6 Sol in five of eight safety categories on its production benchmark and matched or exceeded it on static and multiturn jailbreak defense. On HealthBench Professional it scored 64.2, up 3.4 points from GPT-6 Sol and close to Astra's 64.7. Broken-search-tool disclosure failures fell to 2.1% from 4.9% for GPT-6 Sol.
There are caveats. The addendum reports a misrepresentation rate of 1.50%, versus 0.51% for Astra and 1.30% for GPT-6 Sol, and reports 99.7% on ExploitBench, which it flags at the Critical threshold, against 21.5% on recently disclosed vulnerabilities. Teams with security-sensitive use cases should read the addendum directly.
What to do with it
If you already run agents on GPT-6 Sol, this is a drop-in swap worth testing: same price, better reported results. If you pay for Astra on long coding or computer-use jobs, run a side-by-side on a sample of real tasks and track accepted outputs, retries, and total cost. For latency-sensitive work such as interactive coding or support, wait for the Sol Ultrafast release and measure whether the speed premium pays off.