GPT-4.5
OpenAI's largest-ever non-reasoning model 'Orion' — retired from the API in July 2025.
Retired
GPT-4.5 appears to be discontinued or shut down — Removed from OpenAI API July 14, 2025 (superseded by GPT-4.1); full ChatGPT sunset announced for June 27, 2026. This page is kept as an archive record and is excluded from search engines.
About
GPT-4.5 (codenamed Orion) was OpenAI's largest model by training compute at its release on February 27, 2025 — trained with more compute and data than any prior OpenAI release, with an unsupervised-learning emphasis aimed at world knowledge and emotional intelligence rather than chain-of-thought reasoning. Sam Altman called it "a giant, expensive model" and "the first model that feels like talking to a thoughtful person."
It carried a 128K context window and premium pricing — $75/$150 per million input/output tokens — and OpenAI reported a hallucination rate of 37% versus ~60% for GPT-4o. It never reached frontier-reasoning benchmark levels, and OpenAI pulled it from the API on July 14, 2025 in favor of GPT-4.1; it lingered as a ChatGPT Pro legacy model, with a full ChatGPT sunset announced for June 27, 2026. It remains historically notable as OpenAI's last big pure-scaling release before the reasoning-model era took over.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
Provider-reported: SimpleQA 62.5%, GPQA 71.4%; hallucination rate 37% vs ~60% for GPT-4o.
Independent: no broad third-party reproduction before retirement.
Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.
Use cases
Trust score — 7/10
Release facts, pricing, and retirement timeline corroborated across Wikipedia, TechCrunch, and launch-day press with consistent figures. Marked retired: API access ended July 14, 2025 and full ChatGPT sunset was announced for June 27, 2026. Docked for: (1) all benchmark figures are OpenAI-reported; (2) it is a legacy model — useful as reference, not for new builds.