GPT-4.5

GPT-4.5 is OpenAI’s largest model released before GPT-5, launched on February 27, 2025 for ChatGPT Plus and Pro subscribers and, briefly, through the API. OpenAI described it as trained with a focus on natural conversation and broad world knowledge rather than step-by-step reasoning, and it outperformed GPT-4o on multilingual MMLU testing across the fifteen languages OpenAI checked. Its parameter count was never disclosed, and its context window stayed at 128,000 tokens, unchanged from GPT-4-class models before it.

The model’s defining trait ended up being its price: $75 per million input tokens and $150 per million output tokens, roughly thirty times more expensive than GPT-4o. That cost, combined with GPT-4.1 matching or beating it on coding benchmarks for far less money, led OpenAI to pull GPT-4.5 from the API in mid-2025 and then drop it entirely from ChatGPT Plus and Team tiers when GPT-5 shipped in August 2025, leaving it accessible only to Pro subscribers under a legacy models tab.