- OpenAI has pushed out o3-mini models to ChatGPT
- The launch has previously been teased
- OpenAI is facing increasing competition from China
As promised last week, OpenAI has now launched its latest o3-mini AI model to users on all ChatGPT plans, including the free tier. The new model brings with it improved reasoning capabilities, especially in math, coding, and science.
The o3-mini release “advances the boundaries of what small models can achieve”, OpenAI says, and it apparently responds 24% faster than the o1-mini model it’s replacing. As per external testers, o3-mini answers are preferable to o1-mini answers 56% of the time, and include 39% fewer mistakes.
As with o1-mini, this reasoning AI model will show its workings above its responses – so you can check the ‘thought’ processes involved. You can also combine this reasoning with web searches if needed, though this integration is still in its early stages.
Of course, the release comes after a tumultuous week in AI, in which the models offered for free by China’s DeepSeek have attracted millions of users with their speed and accuracy – and now OpenAI is trying to grab back some of the limelight (and traffic).
Try it yourself
OpenAI o3-mini is now available in ChatGPT and the API.Pro users will have unlimited access to o3-mini and Plus & Team users will have triple the rate limits (vs o1-mini).Free users can try o3-mini in ChatGPT by selecting the Reason button under the message composer.January 31, 2025
Free users can get at o3-mini by clicking the Reason button in the text input box. OpenAI hasn’t specified what the limits on its use will be, but it’s likely to be in line with current restrictions on GPT-4o use – so a handful of queries per hour.
For paying users, o3-mini can be selected from the model picker in the top left corner. If you’re on a Plus or Team plan, you get 150 queries of o3-mini daily, and if you’re on the Pro plan, access is unlimited – for a mere $200 (about £160 / AU$320) per month.
Paying ChatGPT users also get access to an o3-mini-high model that applies the same reasoning skills but takes longer to think and respond. It boosts performance even further, if you don’t mind waiting a few extra seconds.
OpenAI has also highlighted the safety assessments that o3-mini has gone through before launched – it apparently “significantly surpasses” the GPT-4o model when it comes to assessing unsafe use and jailbreak attempts.
You might also like
https://cdn.mos.cms.futurecdn.net/PsDVqkSyiS5mn5Q2evxerN-1200-80.jpg
Source link
