By OpenAI has recently updated ChatGPT with new o3 and o4-mini models, which are now available to Pro, Plus, Team, and even free tier users.
The new models considerably increase ChatGPT’s performance and outperform prior OpenAI reasoning models such as ChatGPT o3-mini and o1.
Most crucially, they may intelligently select from OpenAI’s numerous tools to finish your question, including a new capacity to reason with images.
OpenAI hosted a livestream for the release:
Here are the three major changes:
1. Combining tools.
Both new reasoning models are capable of using and combining every tool in ChatGPT. That means users have full access to ChatGPT’s toolkit, which includes web surfing, Python scripting, picture and file analysis, image production, canvas, automations, file search, and memory.
The crucial point is that ChatGPT now determines if it needs to use a tool depending on what you’ve requested.
When you ask ChatGPT to do something complicated using the new models, it will show you each step it takes, the tool it is using, and how it arrived at that choice.
Once it has completed all of its research, the notes on its working method disappear, and you receive a report on its findings.
2. Better performance
The ability of o3 and o4-mini to intelligently choose which tools to employ is a step towards the intelligent model switching promised by ChatGPT 5, when it is released.
As you would expect from advanced reasoning models, the final report is quite thorough and includes connections to all sources consulted.
OpenAI states: “The combined power of state-of-the-art reasoning with full tool access translates into significantly stronger performance across academic benchmarks and real-world tasks, setting a new standard in both intelligence and usefulness.”
As a result, these models can more successfully address multifaceted problems in the real world, so don’t be hesitant to ask them to conduct multiple actions at once and create an answer or report that integrates multiple queries.
3. Reasoning with images
Both new models are the first from OpenAI that will incorporate uploaded images into its chain of reasoning. They will really reason using the photographs, so you could upload a picture of some automobiles and ask what the name and model are, as well as how much retail value they will have in five year’s time.
This is the first time ChatGPT has been able to incorporate photos into a reasoning chain, representing a significant step forward for multimodal AI.
My verdict:
I tried the new models on the Plus tier and was impressed by the speed and thoroughness of the responses to my questions. While I’ve always appreciated the depth of reasoning afforded by the o1 and o3-mini models, it has always resulted in slower response times.
The o3 model has now been my default model on Plus since it responds quickly enough that I don’t feel like I’m waiting too long for a response, yet it provides a pleasing level of detail.
In simply, I am amazed. The new models appear to be a natural progression of ChatGPT towards more intelligence and capability. I also enjoy how it can determine which of ChatGPT’s different tools it should utilize to deliver the best answer.
Try it for yourself.
Here’s how you may test the new ChatGPT models for yourself:
Plus, Pro and Team customers will notice that they may choose ChatGPT o3, ChatGPT o4-mini, and ChatGPT o4-mini-high from the drop-down LLM menu within ChatGPT, while free tier users can access o4-mini by clicking the Reason button in the composer before submitting their query. Edu users will have access after one week.
There is no visible warning to free tier users that they are now utilising the 4o-mini reasoning model, but if you click the button and ask ChatGPT which LLM it is using, it now says 4o-mini.
There will be a rate limit on how many times a free tier user can utilise the Reasoning feature, while Plus users will have a considerably larger limit.
OpenAI says they plan to release o3-pro “in a few weeks” with complete tool support. Pro users already have access to o1-pro.
Views: 111
