OpenAI o4-mini

o4-mini was released on 16 April 2025, two days after the GPT-4.1 family, as the last small model of the o-series. It kept the shape of o3-mini — a 200,000-token context window, up to 100,000 output tokens, reasoning-token support — and added image input, so a reasoning chain could start from a screenshot or a diagram rather than a description of one. The list price did not move: $1.1 per million input tokens and $4.4 per million output, the same as its predecessor. The one change was cached input, which fell from $0.55 to $0.275 per million, halving the cost of prompts that repeat a long preamble. The knowledge cutoff moved forward to 1 June 2024. OpenAI marketed it for coding and visual tasks and now points developers to GPT-5 mini instead. Two shutdown dates apply. The deep research variant, o4-mini-deep-research, was withdrawn on 23 July 2026, and the model itself, together with its fine-tuned versions, shuts down on 23 October 2026; the recommended replacement is GPT-5.6 Terra.

▸Videos

No videos yet.

▸Reviews