GPT-6.1 Sol Launches: New Sol Version in the API, Codex and ChatGPT Work, Not in Chat

OpenAI launched GPT-6.1 Sol on September 29, 2026, with the API name gpt-6.1-sol. OpenAI's help documentation calls GPT-6 Sol "the previous Sol model," so this is a new version of Sol, but these documents do not say when GPT-6 Sol will be retired; the opening of the system card addendum says "GPT-6.1 is the latest model family in the GPT-6 series." On the ChatGPT side, the launch rollout includes Plus, Pro, Business, Enterprise and Edu in Codex and ChatGPT Work (Enterprise and Edu need an administrator to enable it). It is not in Chat, and Free and Go are not included at launch. Whether accounts in Taiwan can use it is not stated in these documents.

This article draws on OpenAI's API model page, pricing page, the Models page of the ChatGPT and Codex help documentation, and the system card addendum, all checked on September 30, 2026. Evaluations are cited only from the system card addendum. The claims about capability and safety are all published by OpenAI itself; we have not tested the model ourselves, and this is not a recommendation to buy or to switch models.

What GPT-6.1 Sol is, and how it relates to GPT-6 Sol

OpenAI's positioning on the API model page is that GPT-6.1 Sol delivers "near-Astra performance at a lower cost" for complex coding, computer use and professional work. This is OpenAI's own description; it gives no percentage or score gap, and the model page also suggests that developers compare it with Astra on their own tasks to weigh the tradeoff between quality and cost.

The help documentation asks readers to compare GPT-6.1 Sol and GPT-6 Sol themselves to see how their API specifications differ. GPT-6 Sol is still there: the help documentation lists it alongside GPT-6.1 Sol as available in Work and Codex, and the pricing page still lists it as of the check date.

Specifications listed on the API model page (as of the check date): a 1,050,000-token context window, 128,000 max output tokens and an April 30, 2026 knowledge cutoff; input is text and images, and output is text.

Reading the API list prices: next to GPT-6 Sol and GPT-6 Astra

List prices on the API pricing page for Standard, short context (per 1M tokens, US dollars): GPT-6.1 Sol is $2.00 for input, $0.10 for cached input, $2.50 for cache writes and $10.00 for output; GPT-6 Astra is $10.00, $1.00, $12.50 and $50.00. Converting from the list prices, the editors find that GPT-6.1 Sol's input and output list prices are each one-fifth of Astra's.

Compared with GPT-6 Sol: in the All models list on the same pricing page, GPT-6 Sol's short-context list prices are $2.00 for input, $0.20 for cached input, $2.50 for cache writes and $10.00 for output, so only cached input differs ($0.10 versus $0.20).

The GPT-6.1 Sol model page spells out the pricing rules: cached input is priced at 5% of the uncached input rate, and cache writes are billed at 1.25x the uncached input rate, which is why the $2.50 for cache writes is higher than regular input. For prompts with more than 272K input tokens, the whole request is priced at 2x the input and cache rates and 1.5x the output rate. Fast mode prices are 2x Standard, Batch and Flex prices are 50% lower than Standard, and regional processing adds a 10% premium where available.

These are API list prices, not ChatGPT subscription prices; these pages do not state monthly fees for any plan or usage limits for GPT-6.1 Sol.

Who can use GPT-6.1 Sol, and where

The help documentation says that in ChatGPT, GPT-6.1 Sol, GPT-6 Sol and GPT-6 Luna are available in Work and Codex, and not in Chat (the original text is "They aren’t available in Chat"). The documentation does not say whether Chat will get them.

The GPT-6.1 Sol launch rollout includes the Plus, Pro, Business, Enterprise and Edu plans: Codex means the desktop app and CLI, and ChatGPT Work means web and mobile. For Enterprise and Edu, the plan keeps it off by default until an administrator enables it; whether Business also needs an administrator to enable it is not stated in the documentation. Free and Go are not included at launch, and the documentation does not say whether they will be added later.

Standard and Fast modes are available at launch, and Ultrafast support is coming later, with no date given. GPT-5.6 Sol, GPT-5.6 Terra and GPT-5.6 Luna remain available during the rollout. Availability depends on the rollout, your sign-in method and your client; selecting a new model does not change workspace permissions or grant access to it.

As of September 30, 2026, none of these documents states when it will open in Taiwan or other regions, and none lists country or region restrictions for GPT-6.1 Sol.

What the system card addendum says

OpenAI has posted "Addendum to GPT-6 Astra System Card: GPT-6.1 Sol" on its Deployment Safety Hub. It is an addendum to the GPT-6 Astra system card, and the page is dated September 29, 2026.

Under its Preparedness Framework, OpenAI treats GPT-6.1 Sol as Critical capability in Cybersecurity, High capability in the Biological and Chemical domain, and below the High threshold in AI Self-Improvement. The Preparedness Framework is OpenAI's own risk-classification framework; this is an internal classification, not a rating from an outside body.

OpenAI writes that it treats GPT-6.1 Sol as Critical in Cybersecurity, as it did for GPT-6 Astra, and based on this assessment it applied the same safeguards stack as for GPT-6 Astra. As with Astra, OpenAI is taking a phased approach for GPT-6.1 Sol's cybersecurity access through Daybreak; the details are in the GPT-6 Astra system card. OpenAI also says that its Safety Advisory Group's recommendation and OpenAI leadership's determination were that these safeguards are sufficient for the public launch; this is OpenAI's own judgment.

The evaluations in the addendum were performed by OpenAI in its research environment or via its API, which OpenAI notes may provide slightly different output from production ChatGPT. OpenAI says that on its evaluation of user-flagged cases, GPT-6.1 Sol and GPT-6 Sol achieve similarly low hallucination rates, and that GPT-6.1 Sol performs on par with GPT-6 Astra on its HealthBench evaluations. On its coding deception evaluation (Coding Deception), the rate of misrepresentation is 1.50% for GPT-6.1 Sol, 0.51% for GPT-6 Astra and 1.30% for GPT-6 Sol; OpenAI notes that these tasks were deliberately selected to elicit potentially dishonest behavior and that the observed rates are not expected to match the true rate of misbehavior in production.

Checking for yourself: where to look

The most direct way is to open Codex or ChatGPT Work and see whether GPT-6.1 Sol appears in the model picker. If you do not see it, first check the plans and clients above; on Enterprise and Edu accounts, ask your administrator whether it has been enabled.

Developers call it with the name gpt-6.1-sol. The model page says it supports US and EU data residency, Fast mode is unavailable with EU data residency, and billing follows the pricing page on the day of the call.

Prices and rollout scope describe the launch moment and may change later; for the latest status, see OpenAI's API model page, pricing page and the Models page of the help documentation.

2026 AI News Roundup: Highlights and Daily Life Applications from January to September

GPT-6 Sol and Luna Launch: API Price Cuts, Available in Codex, Not Yet in Chat

• AI tools
• Apps & software
• AI news