one year on
OpenAI launches o3-pro reasoning model
The company’s most capable model arrives for ChatGPT Pro and Team subscribers and in the API at $20 per million input tokens and $80 per million output tokens.
Originally published Published on this site
OpenAI today released o3-pro, a more powerful version of its o3 reasoning model, for ChatGPT Pro and Team subscribers. O3-pro replaces o1-pro in the model picker and costs $20 per million input tokens and $80 per million output tokens in the API.
The two moves, announced simultaneously, highlight a tug-of-war in OpenAI’s strategy: pushing the frontier with a premium model while racing to commoditize reasoning capabilities as competitors like Google and Anthropic release their own advanced models.
In internal evaluations, o3-pro outperforms o3 across all tested categories and beats Google’s Gemini 2.5 Pro on the AIME 2024 math benchmark and Anthropic’s Claude 4 Opus on GPQA Diamond, a PhD-level science test. The model supports web search, file analysis, vision, and Python execution. Drawbacks include longer response times than o1-pro, no image generation, and temporary chat being disabled for now.
The record
One year later — open only if you can handle spoilers
Within six months, o3-pro became the default reasoning model for Pro users, but the 80% price cut on o3 reshaped the API market, forcing competitors to slash prices and accelerating the adoption of reasoning models in production applications.
Filed
Compiled and edited by Adam Allcock.
The Weekly Replay · free by email
This week, one year ago — every Sunday.
One email each Sunday: the week's replayed AI news, with the one-year-later annotations included. Written like it's breaking — dated like it isn't.
Free · double opt-in · unsubscribe anytime · privacy