OpenAI reveals updated reasoning models o3 and o4-mini for ChatGPT

UPDATE: Apr. 16, 2025, 2:18 p.m. EDT Now that the OpenAI livestream has concluded, this article has been updated with the latest details regarding the launch of o3 and o4-mini.

OpenAI’s new o3 and o4-mini models for ChatGPT have officially been introduced. As anticipated, during the livestream event earlier today, the creators of ChatGPT unveiled the latest iterations of their reasoning models. Following the launch of o1, their first reasoning model, last summer, there was no announcement regarding o2, but progress on o3 was shared in late December. While o3 has only just become publicly available, OpenAI already had a lightweight version, o3-mini, accessible.

In the livestream, the OpenAI team, minus CEO Sam Altman, highlighted some of the capabilities of these models and demonstrated how users can interact with o3 and o4-mini. These models have been equipped to utilize tools and can process visual content, including charts and images.

“For the first time, our reasoning models can agentically use and combine every tool within ChatGPT—this includes searching the web, analyzing uploaded files and other data with Python, reasoning deeply about visual inputs, and even generating images,” the announcement stated. The o3 model is now recognized as OpenAI’s most sophisticated reasoning model. The streamlined o4-mini is a smaller model aimed at enhancing cost-efficiency, particularly excelling in math and coding, according to OpenAI. Both models outperform o1 and o3-mini in benchmarks such as AIME and GPQA and have shown improvements in visual reasoning based on multimodal evaluations. Initially, in an effort to simplify its array of models, Altman mentioned that they would not release o3 and o4-mini as standalone models and instead would integrate their features into ChatGPT-5. However, a few weeks later, the company opted to release the two models separately, causing yet another delay in the ChatGPT-5 launch. Some observers express concerns about the repeated postponements of GPT-5, suggesting possible internal challenges in enhancing model performance. Nevertheless, Altman assures that the wait will be worthwhile. Currently, o3 and o4-mini are robust enough to function independently. Both models are available today through the API and for ChatGPT Pro, Plus, and Teams subscribers, with access for Enterprise and Educational users expected next week.