GPT-5.1: OpenAI's Adaptive Reasoning Model, No-Reasoning Mode, Developer Tools, Specs and Pricing

What GPT-5.1 is

GPT-5.1 is a model from OpenAI, released on 13 November 2025 as an upgrade to the GPT-5 series. In ChatGPT it arrived as two models: GPT-5.1 Instant, ChatGPT's most-used model, which OpenAI made warmer and more conversational by default, and GPT-5.1 Thinking, its advanced reasoning model, which OpenAI made easier to understand, faster on simple tasks and more persistent on complex ones. In the API, GPT-5.1 Thinking is released as gpt-5.1 and GPT-5.1 Instant as gpt-5.1-chat-latest, both with adaptive reasoning. It is a proprietary model with no published weights. On LMArena's text leaderboard it has 1438 Elo (rank 101).

Adaptive reasoning

OpenAI says GPT-5.1 adapts how long it thinks to the complexity of a task, making it faster and more token-efficient on simple everyday work without degrading performance on hard problems. On a representative distribution of ChatGPT tasks, OpenAI says GPT-5.1 Thinking is roughly twice as fast on the fastest tasks and twice as slow on the slowest. In one example, a request for an npm command took GPT-5 about 250 tokens and 10 seconds at medium effort and GPT-5.1 about 50 tokens and 2 seconds. GPT-5.1 Instant can also use adaptive reasoning for the first time, which OpenAI linked to gains on AIME 2025 and Codeforces. OpenAI quotes Balyasny Asset Management saying GPT-5.1 outperformed GPT-4.1 and GPT-5 in its full evaluation suite while running two to three times faster than GPT-5, and says Priority Processing customers see noticeably faster performance than with GPT-5.

Developer features

Developers can run GPT-5.1 with no reasoning by setting reasoning_effort to none, which is the default; the other options are low, medium and high. Extended prompt caching keeps prompts in the cache for up to 24 hours instead of a few minutes, enabled with prompt_cache_retention set to 24h. OpenAI also introduced an apply_patch tool for more reliable code edits and a shell tool that lets the model run shell commands. On SWE-bench Verified, OpenAI reports GPT-5.1 reaches 76.3%. OpenAI released gpt-5.1-codex and gpt-5.1-codex-mini at the same time for long-running agentic coding.

Specifications and pricing

GPT-5.1 has a 400K-token context window, up to 128,000 output tokens and a knowledge cutoff of 30 September 2024. It takes text and image input and returns text, with reasoning token support. On the API it costs $1.25 per million input tokens and $10 per million output tokens, with cached input at $0.125 per million, a 90% discount, and no charge for cache writes or storage. OpenAI said pricing and rate limits matched GPT-5, and that GPT-5.1 was available on all paid API tiers.

Status

In ChatGPT, GPT-5.1 rolled out first to paid Pro, Plus, Go and Business users and then to free and logged-out users, while GPT-5 stayed in the legacy models menu for paid subscribers for three months. OpenAI's models page lists the gpt-5.1-2025-11-13 snapshot as deprecated, and OpenAI's deprecations page lists GPT-5.1 for shutdown on 1 April 2027, with gpt-6-sol as the replacement.

Official Resources