GPT-4.5 was pitched on warmth and lower hallucination rates rather than reasoning, and it launched as a Pro-tier research preview.
Fewer invented details matters more than raw capability when the output is going in front of a buyer, but a Pro-tier research preview is not something you can standardise a team on.
OpenAI released GPT-4.5 as a research preview on 27 February 2025, available to Pro users and developers worldwide. OpenAI described it as its largest and best model for chat, built by scaling pre-training and post-training rather than by adding reasoning. The company said it expected the model to hallucinate less and to follow user intent more closely.
It is explicitly not a reasoning model, and it shipped behind the Pro tier.
Two minutes, once a week. What changed in AI, and what to run because of it.