Yesterday, OpenAI released the GPT 5.1 series of models for ChatGPT customers. GPT‑5.1 Instant is a warmer, more intelligent, and better at following your instructions, while GPT‑5.1 Thinking is an advanced reasoning model that can perform faster on simple tasks or think longer if required for complex tasks.
Today, OpenAI has expanded the availability of the GPT 5.1 model to its API platform for developers. The main highlight of GPT 5.1 is its adaptive reasoning capability. For straightforward, easy tasks, the GPT‑5.1 model now spends fewer tokens thinking, allowing developers to get an instant response with lower token usage. For complex tasks, GPT‑5.1 will spend more time thinking, exploring options, and providing reliable responses. This new capability will make GPT-5.1 highly efficient for developers.
To make GPT‑5.1 even more efficient, OpenAI is introducing extended prompt caching for up to 24-hour cache retention, which is much better than the few minutes supported today. With prompt caching, developers can not only get faster responses, but they can also reduce the overall API usage cost. OpenAI also highlighted that Priority Processing customers can get noticeably faster performance with this new GPT‑5.1 model over the previous generation.
OpenAI is also introducing a new way for developers to use GPT‑5.1 without reasoning by setting reasoning_effort to "none". Previously, developers who prefer fast responses were forced to use GPT‑5 with "minimal" reasoning. But with this non-reasoning capability, developers can take advantage of the intelligence of GPT‑5.1 with minimal latency.
To improve GPT‑5.1’s coding personality, steerability, and code quality, the OpenAI team worked with leading AI coding startups including Cursor, Cognition, Augment Code, Factory, and Warp.
Finally, developers are getting two new tools with the GPT‑5.1 model. The apply_patch tool will allow the GPT‑5.1 model to create, update, and delete files in a codebase using structured diffs. The new shell tool will allow GPT-5.1 to interact with a local computer through a controlled command-line interface.
Both GPT‑5.1 and gpt-5.1-chat-latest models are available to developers on all paid tiers in the API for the same price as GPT‑5. OpenAI also announced the availability of gpt-5.1-codex and gpt-5.1-codex-mini models that are optimized for long-running, agentic coding tasks in the API.