OpenAI o1-preview
OpenAI · United States · 2024
The first reasoning model anyone could buy — and the moment thinking longer became something you pay for.
o1-preview was released on 12 September 2024 as a research preview, alongside the cheaper o1-mini. It was the first widely available model trained with reinforcement learning to produce a long internal chain of thought before answering: the model spends tokens thinking, those tokens are billed, and the answer improves with the budget. Every reasoning model in this catalogue descends from that pattern. The specification was modest by later standards. Text in, text out — no image input. A 128,000-token context window with up to 32,768 output tokens, a knowledge cutoff of October 2023, and no support for the Responses API, which did not yet exist. Pricing was 15 dollars per million input tokens and 60 dollars per million output, with cached input at half the input rate. Its commercial life was short. OpenAI notified developers of the deprecation on 28 April 2025 and shut the model down three months later, on 28 July 2025, naming o3 as the replacement. o1-mini, announced on the same day, was given six months instead of three. The profile is kept here for the historical record. o1-preview never became a workhorse, but it settled an argument: that inference-time compute is a scaling axis of its own, separate from the size of the model.
▸Videos
No videos yet.
▸Reviews
No reviews yet. Be the first!