Unbelievable: Inside OpenAI’s new o3 and o4 models and what they mean for you

This week in AI has been wild. Some are even claiming we’ve reached AGI, a bold and debatable statement. But what’s clear is OpenAI’s release of “o3” has changed the game, enabling capabilities in ChatGPT that were unthinkable before.

Let’s break down what happened and what it means for regular business users like us.

OpenAI’s big releases: new models and capabilities

OpenAI released several models this week, but I’ll focus on what’s most important for non-developers:

• o3 – The star of the show. This is the powerful “thinking” model that was powering Deep Research before, now available directly in ChatGPT for Pro, Plus and Teams users.
• o4 Mini and o4 Mini High – Smaller versions with different tradeoffs.
• GPT-4.1 series – Developer-focused models (not available in the ChatGPT interface) particularly good at generating code.

What makes o3 special isn’t just that it’s a “thinking model” – it’s that it has access to all ChatGPT’s tools (web search, image analysis, data tools) right out of the gate. Earlier models had limited tool access when first released.

What can o3 actually do? Astonishing examples

The best way to understand o3’s capabilities is through examples. Here are some that showcase its power:

Comprehensive research on people

With a simple prompt like “Do intensive research on [person name] and give me a massive report,” o3 can create detailed profiles that combine information from social media, company websites, publications, and more. The guardrails are lower than Deep Research, which previously refused to research individuals.