GPT-6.1 Sol Costs $2/$10 per Million Tokens With 1.05M Context
OpenAI released GPT-6.1 Sol on September 29, 2026, positioning it between GPT-6 Sol and GPT-6 Astra for complex coding, computer use and professional work. The API model costs $2 per million input tokens, $0.10 per million cached input tokens and $10 per million output tokens at standard rates.
The production model ID is gpt-6.1-sol. OpenAI's current API documentation lists a 1,050,000-token context window, 128,000 maximum output tokens, an April 30, 2026 knowledge cutoff and reasoning-effort settings from low through max. GPT-6.1 Sol is available through the API and in ChatGPT Work and Codex for Plus, Pro, Business, Enterprise and Edu users. OpenAI says it is not yet available in Chat.
For developers choosing between Sol and Astra, the central change is cost per task. OpenAI reports that GPT-6.1 Sol approaches Astra on several agentic and professional evaluations while charging one-fifth of Astra's standard input and output token prices.
GPT-6.1 Sol API specifications
| Specification | GPT-6.1 Sol |
|---|---|
| API model | gpt-6.1-sol |
| Standard input | $2 / 1M tokens |
| Cached input | $0.10 / 1M tokens |
| Standard output | $10 / 1M tokens |
| Context window | 1,050,000 tokens |
| Maximum output | 128,000 tokens |
| Knowledge cutoff | April 30, 2026 |
| Reasoning effort | low, medium, high, xhigh, max |
| Tool-calling API | Responses API |
| Chat Completions | Supported without tool calling |
| Data residency | US and EU supported |
OpenAI's API changelog also lists a $2.50 per million token cache-write price for prompts up to 272K input tokens. Teams using long-lived agents should model cache writes and reads separately because repeated context can shift effective workload cost substantially.
Benchmark results and cost per task
OpenAI reports GPT-6.1 Sol as a substantial step over GPT-6 Sol across software engineering, computer use, scientific workflows and difficult factuality tests.
On DeepSWE v1.1, OpenAI says GPT-6.1 Sol matches GPT-6 Astra at roughly one-fifth of the cost and exceeds GPT-6 Sol's best score by 6.4 percentage points while using a lower reasoning setting.
On Terminal-Bench Science 0.1, GPT-6.1 Sol at maximum reasoning more than doubles GPT-6 Sol's score while costing less than half as much per task. OpenAI reports an average $5.47 per task for GPT-6.1 Sol, compared with $23.80 for GPT-6 Astra and $23.21 for Claude Opus 5.5 in its evaluation. Astra remains the highest-scoring model in that comparison at 68.1%.
OpenAI's difficult-prompt factuality evaluation reports that GPT-6.1 Sol reduces responses containing at least one factual error from 11.4% with GPT-6 Sol to 7.7% at low reasoning effort. OpenAI constructed this evaluation from de-identified conversations in which users had flagged an earlier model error, so the figures describe an intentionally difficult test set rather than typical production error rates.
These are OpenAI-reported evaluations. Workload-specific cost depends on reasoning effort, token volume, cache reuse, tool calls and task completion rate, making cost per successful task more useful than token price alone for production comparisons.
Deployment details
GPT-6.1 Sol supports the Responses API for tool calling. OpenAI's model documentation says Chat Completions is available without tool calling and that none and minimal reasoning efforts are unsupported.
The model supports US and EU data residency. OpenAI notes that Fast mode is unavailable with EU data residency, which is material for deployments that require both regional processing and reduced latency.
OpenAI also lists Multi-agent in beta for GPT-6.1 Sol in the Responses API, allowing the model to delegate work to subagents within a request. This expands the model's role beyond a lower-cost single-agent replacement for Astra and makes context reuse, cache pricing and orchestration overhead more important in capacity planning.
Availability and Ultrafast
GPT-6.1 Sol launched in the API, ChatGPT Work and Codex on September 29. OpenAI says an Ultrafast service tier for GPT-6.1 Sol is planned with up to 8× faster token generation than standard-speed operation in Codex.
OpenAI separately introduced GPT-6 Astra Ultrafast and a $500/month Pro tier with access to the faster service in ChatGPT Work and Codex. The GPT-6.1 Sol launch page describes its own Ultrafast option as coming in the following days, so deployments should use the current API documentation and account availability rather than assuming the faster tier is enabled everywhere.
Where GPT-6.1 Sol fits
GPT-6.1 Sol is most compelling for workloads that need strong coding, computer-use or professional reasoning but run frequently enough for Astra's token prices to dominate operating cost. Its million-token context window and inexpensive cache reads also make it relevant to repository-scale coding agents, document-heavy workflows and persistent agent sessions.
Astra remains OpenAI's higher-capability option for the hardest workloads, including the scientific benchmark where it retains the top published score. For production selection, a representative evaluation should compare successful-task rate, latency and total token/cache consumption at the reasoning settings actually intended for deployment.