GPT-5.1-Codex

GPT-5.1-Codex went out to the Responses API on 13 November 2025, the same day as GPT-5.1 itself and its smaller sibling GPT-5.1-Codex Mini. OpenAI describes it as a version of GPT-5 optimized for agentic coding in Codex or similar environments, which means it is meant to be driven by a coding agent rather than typed into a chat box. The model takes text and images and returns text, with a 400,000-token context window and up to 128,000 output tokens. Knowledge stops on 30 September 2024, more than a year before the model shipped, so it knows nothing about libraries and language releases from 2025; in a coding model that gap matters more than the calendar suggests. Reasoning tokens are supported. Pricing is 1.25 USD per million input tokens, 0.125 USD for cached input and 10 USD per million output tokens. Two restrictions are worth reading before adopting it. Chat Completions is not supported at all - only the Responses API works, and there is no batch route, no fine-tuning and no audio or image generation. And the card states plainly that the underlying model snapshot will be regularly updated: unlike a dated snapshot, the name gpt-5.1-codex points to a moving target, so behaviour can shift underneath an application without any announcement. The model is no longer available. On 22 April 2026 OpenAI announced the withdrawal of a group of older snapshots, and access to gpt-5.1-codex was shut down on 23 July 2026, eight months after release; the substitute named in the deprecation table is GPT-5.6 Sol, a general model rather than a coding-specific one. The documentation page for the model is still online, which makes it easy to mistake a retired product for a current one.

▸Videos

No videos yet.

▸Reviews