one year on
OpenAI launches o3-pro reasoning model and slashes o3 API pricing by 80%
The company's most capable model arrives for top-tier subscribers and developers, while a massive price cut on the base o3 model signals a sudden commoditization of frontier reasoning.
OpenAI today released o3-pro, a more powerful version of its o3 reasoning model, for ChatGPT Pro and Team subscribers. The company also cut the API price of the standard o3 model by 80%, dropping input costs to $2 per million tokens and output to $8 per million tokens.
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. 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.
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.
“This is a big deal for the ecosystem,” one developer posted on social media. “Frontier reasoning just got commodity-priced.” The pricing war on reasoning models, which few expected this early, now places pressure on rivals to match OpenAI’s new rates.
The record
Many expressed surprise at the simultaneous release of a premium model and an 80% price drop on o3, arguing that the pricing changes would make reasoning models viable for startups and indie developers overnight.
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.