Overview
GPT-6.1 Sol was released on September 29, 2026 for complex coding and professional work. OpenAI positions it as offering near-Astra performance at a lower cost than Astra. For people running repeated tasks across code, apps, and documents, it is a candidate to evaluate alongside their existing model. This is OpenAI's positioning, rather than a guarantee that Sol matches Astra on every task. Official release log
This release belongs to the broader ChatGPT model ecosystem, with access in Work, Codex, and the API. Ordinary Chat does not offer GPT-6.1 Sol. The useful upgrade questions are whether your account has access, whether the model handles your workflow well, and whether saved API settings remain compatible. Model availability
What's New
A new Sol option for ongoing Work and Codex tasks
The model ID is gpt-6.1-sol. OpenAI recommends considering it for repeated, long-running work, while retaining GPT-6 Astra for the most demanding workflows. Codex CLI 0.159.1 adds 6.1 Sol as the default in its bundled catalog. A catalog default does not establish what every existing session or saved configuration uses. Product changelog
Lower cached-input pricing
Standard API cached reads fall from GPT-6 Sol's $0.20 to $0.10 per million tokens. Uncached input and output remain $2 and $10. The saving applies to cached input, so a 50% reduction in that rate does not imply a 50% reduction in a task's total bill. API release and previous rates
Different reasoning and tool-call requirements
GPT-6.1 Sol drops the none reasoning option supported by GPT-6 Sol. It also requires the Responses API for tool calling; Chat Completions works for requests without tools. An application using GPT-6 Sol with Chat Completions function calls and none effort needs a migration beyond changing the model ID. New model, previous model
The launch also documents beta multi-agent support in Responses. Treat this as documented support for 6.1 Sol, rather than a claim that this release first invented delegation across the GPT family. API changelog
Compared With Previous Version
The direct predecessor is GPT-6 Sol, introduced with Luna on September 22. The earlier Sol and Luna release explains the family and its lower-cost tiers; this page focuses on the subsequent Sol update.
| Decision factor | GPT-6 Sol | GPT-6.1 Sol |
|---|---|---|
| API model ID | gpt-6-sol |
gpt-6.1-sol |
| Standard input / output per 1M tokens | $2 / $10 | $2 / $10 |
| Standard cached input per 1M tokens | $0.20 | $0.10 |
| API reasoning effort | none, low, medium, high, xhigh, max |
low, medium, high, xhigh, max |
| Chat Completions tool calling | Function calling only at none effort |
Not supported; use Responses |
| Context / maximum output | 1.05M / 128K tokens | 1.05M / 128K tokens |
Specifications and previous rates come from the GPT-6 Sol model page; current rates are listed in API pricing. Context capacity is unchanged. Evaluate completion quality, retries, elapsed time, and total usage on representative tasks before selecting a default; the capacity numbers alone do not establish a performance gain.
Availability and Access
Launch rollout includes Plus, Pro, Business, Enterprise, and Edu in Codex in the desktop app and CLI, and ChatGPT Work on web and mobile. Enterprise and Edu administrators must enable the model; it is off by default there. Free and Go are excluded at launch. Availability also depends on client and workspace settings. Official availability guide
In a compatible Codex CLI, select it through /model or start with codex --model gpt-6.1-sol. Start with the reasoning effort available by default in your client and increase it when the task warrants deeper analysis. Standard and Fast modes are available at launch; Ultrafast support for 6.1 Sol is planned for later. These product controls and entitlements should be checked separately from API parameters and token rates. Model selection and rollout
API Costs and Migration
Standard API rates, per million tokens for requests with up to 272K input tokens, are $2 input, $0.10 cached input, $2.50 cache writes, and $10 output. Above that input threshold, the long-context rates apply to the full request. API token pricing is a separate billing basis from a ChatGPT subscription or product credit allowance. API pricing
For an existing GPT-6 Sol integration:
- Select
gpt-6.1-soland preserve the existing reasoning effort where supported. Replacenonewithlowand evaluate the result. - Move tool calls to Responses if the application currently uses Chat Completions. Text requests without tools can still use Chat Completions.
- Remove unsupported
temperature,top_p, and log-probability options when using reasoning, following the endpoint-specific guidance. - Compare representative runs before updating saved defaults. Track completion quality and total usage as well as cached-read savings.
The first three steps follow OpenAI's migration guidance; the evaluation step is an operational recommendation.
Who Should Upgrade / Who Should Wait
Evaluate 6.1 Sol now if you have Work or Codex access and repeatedly complete coding or document tasks. It is also worth testing when cached API input is a significant part of your costs. Use a familiar task with explicit acceptance criteria so the choice reflects completed work, rather than the model name alone.
Keep Astra for your hardest workflows when its stronger reasoning and judgment are essential. OpenAI still positions Astra as its most capable model; test any proposed substitution against your requirements. GPT-6 model guidance
Wait before switching if your workspace has not enabled 6.1 Sol, you use ordinary Chat, or your integration depends on none reasoning or Chat Completions tool calls. Resolve access or compatibility first, then evaluate the new model.
