codex-mini-latest
OpenAI · USA · 2025
OpenAI's first Codex CLI model: a fine-tune of o4-mini that cost more than the model it was built from, and was switched off nine months later.
codex-mini-latest was launched in the API on 15 May 2025, a month after OpenAI released the Codex CLI itself, and it is the first model the company shipped specifically for that tool. Technically it is not a new model at all: the card describes it as a fine-tuned version of o4-mini for use in Codex CLI, and adds an unusual piece of advice - for direct use in the API, OpenAI recommends starting with GPT-4.1 instead. The specification is modest next to the later Codex models: a 200,000-token context window, up to 100,000 output tokens, a knowledge cutoff of 1 June 2024, text and image input, text output and reasoning token support. Pricing was 1.5 USD per million input tokens, 0.375 USD cached and 6 USD per million output. The card's own comparison table makes the awkward point for us: o4-mini, the model this was fine-tuned from, cost 1.1 and 4.4 USD - so the Codex variant was roughly a third more expensive than its own base. OpenAI notified developers of its retirement on 17 November 2025 and removed it from the API on 12 February 2026, pointing users to gpt-5-codex-mini. The same notice ended support for the legacy local shell tool, which was only ever available with this model. One detail of its history is only visible in the changelog: at launch the model was offered through both the Responses API and Chat Completions, but the model card as published later lists Chat Completions as not supported - the access route narrowed while the model was live.
▸Videos
No videos yet.
▸Reviews
No reviews yet. Be the first!