OpenAI rolls out new o3-pro AI model to ChatGPT Pro subscribers

OpenAI has launched o3-pro, a high-compute AI model designed to deliver more reliable answers, especially for math, science, and coding. It replaces o1-pro in the ChatGPT Pro plan starting today. So what’s new?

Well, o3-pro has full tool support. That means the model can tap into things like Python, file uploads, memory, visual inputs, and even web search when the prompt demands it. However, it takes a bit longer to respond. OpenAI says the delay is worth it if you need “reliability over speed.”

OpenAI’s new o3-pro model is now live for ChatGPT Pro users

The new model is available through the model picker in both the Pro and Team tiers. Enterprise and Education customers are set to get access next week. Developers can already start using it via API, with pricing set at $20 per million input tokens and $80 per million output tokens.

It also supports a 200,000-token context window, with OpenAI suggesting the use of background mode to avoid timeouts. According to OpenAI’s internal tests, users prefer o3-pro over the standard o3 model in categories like education, science, business writing, and software development. It also scored higher in clarity, accuracy, and instruction-following.

Also read: OpenAI Reportedly Strikes Deal with Google Cloud to Ease ChatGPT Server Load

Not all features are live just yet

A few features aren’t live yet. Temporary chats are disabled due to a technical issue, and image generation isn’t supported within o3-pro—you’ll need to use GPT-4o or other models for that.

Canvas, a tool used to collaborate on long-form content or code, also isn’t yet compatible. It’s worth noting that the o3-pro model was last trained on data up to May 31, 2024. Moreover, OpenAI also slashed the price of the o3 model by 80% to compete directly with Google’s Gemini 2.5 Pro.