OpenAI introduces GPT-6.1 Sol with stronger performance at lower cost

OpenAI introduced GPT-6.1 Sol on September 29, 2026, as an upgrade to GPT-6 Sol. The company positions it as a model for coding, computer use and professional work that approaches GPT-6 Astra on several evaluations at a fraction of the price. OpenAI says the new model is available from launch to Plus, Pro, Business, Enterprise and Edu subscribers in ChatGPT Work and Codex. It is not yet available in Chat, while developers can call it through the API as gpt-6.1-sol.

The listed API price is $2 per million input tokens and $10 per million output tokens. Cached input costs $0.10 per million tokens. OpenAI describes that cached rate as 95% below its standard input price and half the cached-input price of GPT-6 Sol. The company says GPT-6.1 Sol’s standard input and output rates are about one-fifth of GPT-6 Astra’s. These prices matter for developers running agents repeatedly, where a model may reuse a large prompt or work context across many requests. OpenAI says an Ultrafast option, with up to eight times faster token generation in Codex, will follow in the coming days.

OpenAI reports gains on a range of evaluations. On DeepSWE v1.1, a software-engineering benchmark, GPT-6.1 Sol matches Astra while costing about one-fifth as much, according to the company. On OSWorld 2.0 computer-use tasks it improves seven percentage points over GPT-6 Sol at maximum reasoning effort and comes within 2.1 points of Astra. OpenAI also reports better results on complex PDF work, multi-step business workflows, scientific tasks and factuality prompts. The company cautions that some tests use deliberately difficult or specialized tasks and do not represent typical use. Its benchmark results are vendor-reported; performance can also vary with tools, prompts and reasoning settings.

For users, the release offers a more capable option inside Work and Codex without Astra’s listed price. Lower costs may make longer coding and business workflows practical to run more often. However, access depends on plan and product, and API teams should compare the model on their own workloads before switching. GPT-6.1 Sol’s release is another move toward offering several capability and price tiers, giving buyers more room to balance quality, speed and operating cost.