OPENAI

o3-pro showed up with receipts and reasoning

OpenAI has launched o3-pro, a new AI model built to handle complex things like maths, physics, coding, and more.

It’s an upgrade to o3, which came out earlier this year, and promises better accuracy and clearer responses.

It’s now available to ChatGPT Pro and Team users, replacing the older o1-pro. Enterprise and Education users will get access next week.

Developers can also use it right away through OpenAI’s API.

The API pricing is $20 per million input tokens and $80 per million output tokens. (That’s around 750,000 words per million input tokens.)

According to OpenAI, internal tests show:

  • o3-pro performs better than o3 in science, writing, education, and business tasks.

  • It outscored Google’s Gemini 2.5 Pro in maths, and beat Claude 4 Opus in a PhD-level science test.

Pay to think, literally

The model can use tools like web search, file analysis, Python, and memory for personalised replies.

That said, it can’t generate images, doesn’t work with Canvas yet, and temporary chats are switched off while they fix the technical issue.

It also runs a bit slower than o1-pro.

o3-pro can’t draw, but it can probably explain quantum mechanics better than your uni professor.

Keep Reading

No posts found