- After DeepSeek released its R1 model for free, OpenAI has made its frontier 'o3-mini' model freely available to all ChatGPT users.
- ChatGPT Plus users get the 'o3-mini-high' model which is even more capable in coding, science, math, and reasoning tasks.
- Both models are rolling out today. Free ChatGPT users can select "Reason" in message composer to chat with the o3-mini model.
The Chinese AI lab, DeepSeek, recently released its o1-level reasoning model called R1 and made it freely available to all users globally. This unexpected development shocked the US tech stock market and pushed OpenAI to release its frontier AI models. Finally, today, OpenAI has released the powerful ‘o3-mini’ model for free to all ChatGPT users.
Earlier, OpenAI was serving the much smaller “GPT-4o mini” to free ChatGPT users, with some GPT-4o usage as well. Now, all ChatGPT users can access the ‘o3-mini’ model which was announced in December, last year. The o3 series models were hailed as a breakthrough because o3 at the highest compute setting, cracked the hallowed ARC-AGI benchmark.
Talking specifically about the ‘o3-mini’ model which will be available to free users, its reasoning effort is set to “medium”. In AIME 2024 (a competitive math benchmark), o3-mini nearly matches the performance of the much larger o1 model.
In Codeforces (a competitive coding benchmark), o3-mini again performs better than o1. In terms of latency too, o3-mini is almost on par with GPT-4o. OpenAI says the o3-mini model delivers exceptional performance in science, math, coding, and reasoning problems.
As for ChatGPT Plus users, OpenAI has released the ‘o3-mini-high’ model which is even more powerful and the most capable AI model out there for coding. o3-mini-high scores much better than o1 in math and coding benchmarks including AIME 2024, Codeforces, and SWE-bench Verified. Only in GPQA Diamond (PhD-level Science Questions), o1 performs better than o3-mini-high.
Both o3-mini and o3-mini-high models are rolling out to free and paid ChatGPT users, starting today. Of course, there is a rate limit for free users, but OpenAI has not disclosed it. However, OpenAI has tripled the rate limit for ChatGPT Plus users.
If you are a paid subscriber, the rate limit has been increased to 150 messages per day with o3-mini. Note that ChatGPT Pro users who are subscribed to the $200-per-month plan will have unlimited access to both o3-mini and o3-mini-high models.
All in all, for the first time, free ChatGPT users will have access to a far superior model, thanks to DeepSeek’s entry into the AI race. To use the o3-mini model, just select the “Reason” button in ChatGPT and start asking your questions.
