OpenAI's reasoning models just got two useful updates


ZDNET

OpenAI’s reasoning models, o1 and o3-mini, are helpful tools for people who rely on ChatGPT for more complex prompts, including coding, math, and even multi-step text prompts, such as working through a problem or building an itinerary. To make the experience more helpful, OpenAI has deployed small but mighty updates. 

Also: 3 easy side hustles OpenAI’s Operator just made possible – plus how you can get started

OpenAI announced via an X post that OpenAI o1 and o3-mini now support file and image uploads in ChatGPT. The ability to upload images is one of the most powerful ChatGPT features because it allows the chatbot to assist as you work with everyday resources, whether a photo, document, graph, or PDF:

Combining the ability to upload files and the advanced reasoning of ChatGPT opens a new range of use cases. For example, when uploading a long research paper, ChatGPT can provide a more thorough and meaningful analysis than before. Another use case is a complex math problem you are trying to solve on paper. Now, you can upload a photo and get accurate assistance. 

The best part is that, unlike when most new features are first limited to paid subscribers, free users can access the o3-mini model with uploads. 

Also: Will AI kill creativity and destroy human innovation? Here’s the unsettling truth

As a free user, click “Reason” in the message textbox and attach a photo or document to your prompt. Paid users can select the model from the model picker and upload an image as they would regularly. 

In the announcement, OpenAI also raised the usage limits for o3-mini-high, making it seven times greater for Plus users with a capacity of up to 50 messages per day. 

Also: Who should run OpenAI? 50% say Altman and only 23% say Musk, in ZDNET poll

The o3-mini model with high-reasoning effort performed significantly better than o3-mini in the benchmarks OpenAI posted when it first released the model, coming close to o1’s performance, sometimes even surpassing it, as seen in the AIME 2024





Source link

Leave a Comment