GPT-5.1-Codex Mini

GPT-5.1-Codex Mini was released to the Responses API on 13 November 2025, the same day as GPT-5.1-Codex. OpenAI introduces it in one sentence and does not soften it: a smaller, more cost-effective, less-capable version of GPT-5.1-Codex. There is no benchmark table on the card to say how much less capable. What the card does make easy is the arithmetic. Input costs 0.25 USD per million tokens, cached input 0.025 USD and output 2 USD per million - exactly one fifth of the full model's 1.25 and 10 USD, on every line. Everything else is held constant: the same 400,000-token context window, the same 128,000-token output limit, the same 30 September 2024 knowledge cutoff, the same text-and-image input, the same reasoning token support. So the choice between the two is a pure quality-for-money trade with no hidden differences in the specification. One small inconsistency is worth flagging, because it changes what an agent can do: web search appears in the card's list of supported tools but is missing from the list of supported features, where the full GPT-5.1-Codex has it. As with the rest of the line, only the Responses API is supported - Chat Completions, batch, fine-tuning and every audio or image route are closed. The model is no longer available. Access was shut down on 23 July 2026, in the same group withdrawal announced on 22 April 2026 that took the whole GPT-5 and GPT-5.1 Codex family. The substitute named for the Mini version is GPT-5.6 Terra - a different replacement from the one offered for the larger Codex models, which were pointed at GPT-5.6 Sol.

▸Videos

No videos yet.

▸Reviews