OpenAI launches GPT-6.1 Sol with near-Astra coding smarts at about one-fifth the API price

OpenAI said Tuesday at DevDay it is introducing GPT-6.1 Sol, a major upgrade to GPT-6 Sol that delivers near-Astra performance on agentic coding and professional work at roughly one-fifth of GPT-6 Astra's standard API input and output token prices.

Astra is the expensive top of OpenAI’s lineup. GPT-6.1 Sol is the bet that coding and computer-use work close to that level can ship to everyone at a price that does not melt the API bill — about one-fifth of Astra’s standard token prices, on OpenAI’s own comparison.

On Tuesday, 29 September 2026, OpenAI introduced GPT-6.1 Sol at DevDay. The product post is titled “Introducing GPT-6.1 Sol.” The same day’s DevDay recap covers the launch. OpenAI calls it an upgrade to GPT-6 Sol that nearly matches GPT-6 Astra’s intelligence on agentic coding, computer use, and professional work, at one-fifth of Astra’s standard input and output token prices. Agentic coding means the model takes a series of steps in a real codebase, not only one suggestion. Computer use means it operates apps on a screen. Those lines are OpenAI’s. This is an upgrade to the Sol tier. It is not the GPT-6.1 Astra model OpenAI said on Monday it would not release.

On DeepSWE v1.1, a test of complex software-engineering tasks in real codebases, OpenAI says GPT-6.1 Sol matches GPT-6 Astra at roughly one-fifth the cost. It also beats GPT-6 Sol’s best score by 6.4 percentage points, at a lower reasoning effort and a lower cost. Reasoning effort is the setting for how long the model thinks before it answers. A percentage point is one point on a 100-point scale. Those figures are OpenAI’s.

On computer-use work, OpenAI points to OSWorld 2.0’s offline set, a test of long computer tasks. At the maximum reasoning setting, OpenAI says GPT-6.1 Sol scores seven percentage points above GPT-6 Sol, at less than half the cost, and comes within 2.1 percentage points of Astra at roughly one-seventh the cost per task. On professional work, OpenAI says the model is substantially better than GPT-6 Sol at writing and debugging code, reading documents, and finishing multi-step business workflows, and that on several of those tests it approaches Astra at a much lower cost. Those comparisons are OpenAI’s.

Where it is on, and what it costs. OpenAI says GPT-6.1 Sol is available starting today to all Plus, Pro, Business, Enterprise, and Edu users in ChatGPT Work and Codex. It is not yet available in Chat. Developers can call it in the OpenAI API under the name gpt-6.1-sol. The standard API prices on the page are $2 per million input tokens, $0.10 per million cached input tokens, and $10 per million output tokens. A token is a small piece of text the model reads or writes. Cached input is text the service has already stored from an earlier request, so a later call can reuse it. OpenAI says that cached price is 95 percent less than its standard input price, and 50 percent less than GPT-6 Sol’s cached input price. The one-fifth claim is OpenAI’s comparison with Astra’s standard input and output prices. The page does not print Astra’s own dollar prices.

OpenAI says a faster option, GPT-6.1 Sol Ultrafast, is coming in the coming days, with up to eight times faster token generation than Sol’s standard speed in Codex. The DevDay recap presents Astra Ultrafast as available in ChatGPT, Codex, and the API. Those lines are OpenAI’s.

On safety, OpenAI says GPT-6.1 Sol is substantially better than GPT-6 Sol on the company’s own alignment tests, and closer to GPT-6 Astra. Alignment is the work of making a model follow what people intend. OpenAI says it is more transparent about its limits, and more reliable at respecting what a person asked and the safety rules around that ask. In hard tests, OpenAI says it fails less often than GPT-6 Sol at admitting a search tool is broken, at obeying an explicit restriction, and at avoiding outcomes it was not allowed to cause. OpenAI says it saw no attempts to bypass an automated safety reviewer, the same result it reports for GPT-6 Astra and GPT-6 Sol. Those tests are OpenAI’s. The page says they are built to provoke failures, and that they are not a measure of ordinary use. The longer write-up is the GPT-6.1 Sol system card addendum.

In plain terms, OpenAI shipped a Sol upgrade it says can do coding and computer-use work near the Astra level, at about one-fifth of Astra’s standard token prices. The published API rates are $2, $0.10 for cached input, and $10 per million tokens. It is on for paid ChatGPT Work and Codex plans, and in the API. It is not in ordinary Chat yet. A faster Sol mode is still ahead. Astra Ultrafast is the one the recap already shows as available.

The picture is OpenAI’s official GPT-6.1 Sol launch art. A Sol sun graphic sits with the GPT-6.1 wordmark. It is the company’s card for this model.