OpenAI releases GPT-6.1 Sol for complex coding and professional work at a fifth of Astra's price

On September 29th, OpenAI released GPT-6.1 Sol, a model for complex coding, computer use and professional work. It's in the API, through the Responses and Chat Completions endpoints.

Slide text

01 02 03 04 What it is

OpenAI API changelog · Sep 29, 2026 · OpenAI API docs · GPT-6.1 Sol model page

OpenAI releases GPT-6.1 Sol

September 29, 2026 Complex coding and professional work

What the model is.

OpenAI describes GPT-6.1 Sol as near-Astra performance at a lower cost, which is OpenAI's own claim. Its advice is to run both on your own tasks and weigh quality against price.

It takes the middle spot that GPT-6 Sol held a week ago. Astra is for the hardest work, Sol balances speed, cost and intelligence, and Luna is the fast, cheap one.

It takes text and images and writes text. The context window holds about a million tokens, one reply can run to 128,000 tokens, and its knowledge stops at April 30th, 2026.

What it costs.

Per million tokens, GPT-6.1 Sol costs 2 dollars in and 10 dollars out. Astra costs 10 and 50, so OpenAI puts Sol at one fifth of Astra's standard prices.

The list prices match GPT-6 Sol, but repeated context got cheaper. Cached input was 20 cents per million tokens. On GPT-6.1 Sol it's 10 cents, 5 percent of the normal input price.

Batch and Flex processing cost half. Fast mode costs double. And a prompt over 272,000 input tokens doubles the input price and adds half to the output price, for the whole request.

Multi-agent, in beta.

GPT-6.1 Sol supports Multi-agent, in beta. In one Responses API request, the model spins up subagents, runs them in parallel, and combines their work into one answer. Among GPT-6 models, the guide lists only Sol.

OpenAI suggests it for work that splits cleanly, like separate parts of a big codebase. It helps less when each step depends on the last. Subagents can raise token use, and the beta's schemas may change.

Who can use it, and how.

On the API's free tier, GPT-6.1 Sol isn't supported. Usage tier 1 starts at 500 requests a minute. Tier 5 reaches 15,000 requests and 40 million tokens a minute.

GPT-6.1 Sol supports US and EU data residency, but Fast mode is unavailable with EU residency. Regional processing adds 10 percent to the price where it's available.

For tool calling, use the Responses API, where Sol gets web search, file search, computer use, MCP and more. Chat Completions works only without tools. GPT-6 Sol allowed function calls there with reasoning turned off.

Reasoning effort runs from low, through medium, the default, up to max. The none and minimal settings aren't supported. On GPT-6 Sol, none was available.

If you already use GPT-6 Sol, OpenAI asks you to review its migration guidance before switching. It's in the Using GPT-6 guide, on OpenAI's developer site.