GPT-5
OpenAI · USA · 2025
The August 2025 flagship that introduced the router — one name hiding a fast model, a reasoning model and a decision about which to use. Switched off on 11 December 2026.
GPT-5 launched on 7 August 2025 and was treated at the time as a generational landmark. Its most consequential idea was not a benchmark score but a change in what "a model" means: in ChatGPT, GPT-5 was a system rather than a single network — a fast model, a slower reasoning model, and a router that decided which one should answer. Users stopped picking a model from a dropdown; the product picked for them. The API told a different story. There OpenAI shipped three separately sized reasoning models — gpt-5, gpt-5-mini and gpt-5-nano — with an explicit reasoning-effort control at minimal, low, medium and high. The flagship snapshot gpt-5-2025-08-07 carries a 400,000-token context window (up to 272,000 input and 128,000 output), a knowledge cutoff of 30 September 2024, and list prices of $1.25 per million input tokens, $0.125 cached and $10 output. The smaller tiers came in at $0.25/$2 (mini) and $0.05/$0.40 (nano). Streaming, structured outputs, function calling, prompt caching, web search, file search, image generation, code interpreter and MCP were all supported. At launch OpenAI reported 94.6% on AIME 2025 without tools, 74.9% on its 477-task run of SWE-bench Verified, 88.0% on Aider Polyglot, 84.2% on MMMU and 46.2% on HealthBench Hard at high reasoning effort. The often-quoted 88.4% on GPQA Diamond without tools belongs to GPT-5 Pro, not to the base model — a distinction worth keeping straight when comparing generations. GPT-5's service life turned out to be about sixteen months. OpenAI's deprecation schedule shuts down gpt-5-2025-08-07 on 11 December 2026, with gpt-5.6-sol named as the recommended replacement. In the meantime the family moved through five point releases; the profiles for GPT-5.5 and GPT-5.6 cover where the line went next.
▸News
▸Videos
No videos yet.
▸Reviews
No reviews yet. Be the first!