GPT-5 mini

GPT-5 mini launched on 7 August 2025 alongside GPT-5 and GPT-5 nano as the middle tier of the first GPT-5 wave: near-frontier quality for cost-sensitive, low-latency, high-volume work. It shares the flagship's 400,000-token context window, 128,000-token output limit, text and image input, and reasoning-token support, and carries a knowledge cutoff of 31 May 2024. The economics are the point. At $0.25 per million input tokens and $2 per million output, it costs a fifth of what GPT-5 charged at launch, and cached input drops to $0.025 — a hundredth of the flagship's input rate. For well-defined tasks with precise prompts, that ratio, not the benchmark table, is what decided which model went into production. OpenAI announced the end of the line on 11 June 2026: gpt-5-mini shuts down on 11 December 2026, together with GPT-5, GPT-5 nano, GPT-5 Pro, o3 and o3-pro. The recommended replacement is GPT-5.6 Terra. Both the older model and its successor have their own entries in this catalogue.

▸Videos

No videos yet.

▸Reviews