GPT-5 Pro

GPT-5 Pro, released on 6 October 2025 at OpenAI's developer day, is the high-compute version of GPT-5: the same model family asked to think longer before answering. It takes text and images, returns text, and keeps the 400,000-token context window — but raises the output ceiling to 272,000 tokens, more than double the 128,000 of every other model in the generation. It is deliberately narrow in how it can be used. GPT-5 Pro runs in the Responses API only, which lets the model take multiple internal turns before replying, and OpenAI warns plainly that some requests may take several minutes. Background mode or streaming is the recommended way to call it. Pricing reflects the compute: $15 per million input tokens and $120 per million output, twelve times the GPT-5 launch rate on both sides, with no cached-input discount published. The knowledge cutoff is 30 September 2024. OpenAI announced the shutdown on 11 June 2026: the model switches off on 11 December 2026, and the recommended replacement is GPT-5.6 Sol run with reasoning.mode set to pro — the successor generation folds the Pro tier back into the base model as a parameter instead of a separate product.

▸Videos

No videos yet.

▸Reviews