Last week, OpenAI unveiled the o1 model globally. This advanced model is designed to handle complex tasks in mathematics, coding, and intricate problem-solving. The company has indicated that future updates will enhance its ability to address PhD-level scientific inquiries, providing solutions to highly specialized scientific questions in the future. Alongside the o1 model, OpenAI also released the o1-mini. As its name suggests, the o1-mini is more affordable and offers limited features compared to the preview version. This mini version is a faster and cheaper reasoning model, particularly effective for coding tasks.
The o1-mini model has comparable reasoning capabilities, making it a cost-effective choice for applications that prioritize logical reasoning over extensive knowledge. While ChatGPT Plus and Team users can access the o1 models within ChatGPT, the good news is that the o1-mini is available to everyone for free. Both o1-preview and o1-mini can be manually selected in the model picker, with initial weekly limits set at 30 messages for the o1-preview and 50 for the o1-mini. Here's how to use the free version.
Using OpenAI o1-miniIt's important to note that the OpenAI o1 model is being rolled out gradually, so it may take some time before it becomes available for you. To access o1-mini, users should open ChatGPT on their desktop and start a conversation.
Once you've done that, go to the top of the window, click on ChatGPT Auto, and select o1-mini from the Alpha Models dropdown. After this selection, you'll notice that the ChatGPT Auto text has changed to ChatGPT Alpha.
To switch models on the mobile app, long press a ChatGPT response and tap Change model at the bottom of the context menu. You will see the Alpha [gpt-4o] model; select this to begin using the o1-mini version.
What’s Missing in OpenAI o1?OpenAI has confirmed that the new o1 model lacks several features found in the ChatGPT 4o model. This initial version has certain limitations: it cannot browse the internet for problem-solving and does not allow users to upload images. Consequently, it can currently only process text prompts. So, if you were hoping to upload a picture for assistance, that feature isn't available yet.
In summary, the OpenAI o1 model is designed to help users tackle complex math and academic problems requiring reasoning.
The company has also mentioned that these models have been trained to spend more time thinking through problems before responding, similar to human thought processes. Through training, the models have learned to refine their reasoning, experiment with different strategies, and recognize their mistakes.