GPT-5-Codex
OpenAI · USA · 2025
The model that started OpenAI's separate coding line, with a 400,000-token window capped at 272,000 tokens of input. Shut down on 23 July 2026.
GPT-5-Codex is the model that split OpenAI's coding work off into its own product line. The changelog records it as a special-purpose model launched on 23 September 2025, built and optimized for use with the Codex CLI; everything later in the family - the GPT-5.1, 5.2 and 5.3 Codex models - follows the pattern it set. The card gives it a 400,000-token context window with an explicitly stated ceiling of 272,000 input tokens and 128,000 output tokens, text and image input, text output, and reasoning token support. Knowledge stops on 30 September 2024. Pricing is 1.25 USD per million input tokens, 0.125 USD cached and 10 USD per million output - the rate that held for the whole GPT-5 and GPT-5.1 Codex generation and only rose with GPT-5.2-Codex, to 1.75 and 14 USD. The access model was set here as well, and it is unusually narrow: Responses API only, no Chat Completions, no batch, no fine-tuning, no embeddings and no audio or image routes. Supported tools are function calling and web search. As with the rest of the line, gpt-5-codex is a rolling snapshot rather than a dated one - OpenAI states that the underlying model will be updated regularly under the same name. The model is no longer available. Access was shut down on 23 July 2026, ten months after launch, in the group withdrawal announced on 22 April 2026, with GPT-5.6 Sol named as the substitute. That single shutdown removed the entire first half of the Codex line - gpt-5-codex, gpt-5.1-codex, gpt-5.1-codex-max, gpt-5.1-codex-mini and gpt-5.2-codex went the same day - leaving GPT-5.3-Codex as the oldest coding model still reachable through the API.
▸News
▸Videos
No videos yet.
▸Reviews
No reviews yet. Be the first!