OpenAI has launched two new AI reasoning models. These models are called o3 and o4-mini. They are designed to think carefully before giving an answer. All of these models are now available to users with Pro, Plus, or Team plans in ChatGPT.
OpenAI says that O3 is its best reasoning model so far. It performs better than past models in math, coding, science, and understanding images. The o4-mini model is made for developers who want a balance between price, speed, and performance.
These new models can use extra tools inside ChatGPT. They can browse the web, write and run Python code, understand images, and even create images. There is also a version of o4-mini called o4-mini-high. It takes a bit more time to answer, but it gives better results.
OpenAI is working hard to stay ahead in the AI race. Other companies like Google, Meta, and Anthropic are also making advanced models. OpenAI was first to launch a reasoning model (called o1), but others soon followed.
At first, OpenAI did not plan to release o3 in ChatGPT. CEO Sam Altman had said they were working on a better model that uses O3’s technology. However, strong competition pushed them to release O3 anyway.
The O3 model scored 69.1% on a coding test called SWE-bench. This is better than many other models. The o4-mini scored 68.1%, and the older o3-mini scored just 49.3%. For comparison, Anthropic’s Claude 3.7 Sonnet scored 62.3%.
OpenAI also says that O3 and O4-mini are their first models that can “think with images.” You can upload images like sketches or diagrams, and the models will understand them. They can even look at blurry images, zoom in, or rotate them while thinking.
Both models can also run Python code inside your browser using ChatGPT’s Canvas tool. They can also search the web when asked about current events.
Developers can access o3, o4-mini, and o4-mini-high using OpenAI’s APIs. This lets them build apps using these models.
OpenAI has set a low price for the o3 model. It costs $10 per million input tokens and $40 per million output tokens. The o4-mini costs the same as o3-mini: $1.10 per million input tokens and $4.40 per million output tokens.
Soon, OpenAI will also launch o3-pro, a more powerful version of o3. It will be available only to ChatGPT Pro users.
CEO Sam Altman says these may be the last standalone reasoning models before OpenAI releases GPT-5. GPT-5 is expected to combine everything into one advanced model.
