ChatGPT o1 model briefly escapes preview mode

OpenAI accidentally released the full ChatGPT o1 model to the AI chatbot platform a few days ago. The leak only lasted a few hours, but that was plenty of time to show that the model will be even more formidable than the genius riddle solver preview version introduced to the platform in September.

ChatGPT o1’s advances from the current standard ChatGPT-4o model are undeniable, even as a preview. It’s more accurate, better at solving complex, multi-step problems, and can write advanced software code. However, the preview doesn’t include all of the features available in the default ChatGPT-4o, like data analysis, visual understanding, and the ability to search the web.

Those tools were all there in the final model before OpenAI yanked it. In those few hours, users showcased how ChatGPT o1 solved visual puzzles, described photographs in minute and accurate detail, and could break down a text conversation from a screenshot, even replicating the emojis used. You can see some examples below.

ChatGPT o1 thinks slow but deep

The big difference with ChatGPT o1 compared to its predecessors is that it is supposed to use reason and logic to break down questions and tasks into smaller parts and solve each of them before responding. That should make it both less likely to hallucinate and more capable of putting its response into the context of the query. OpenAI hasn’t made any announcements about specific release dates for ChatGPT o1, but the more public tests, even the inadvertent ones, suggest it will happen very soon, maybe before the end of the year.

There’s plenty of reason to think it will be a boon for OpenAI as ChatGPT faces ever more advanced competition from the likes of Google Gemini, Meta AI, and other AI chatbots. The reduced response speed is very much overcome by the more accurate and useful answers provided when asking about either text or images. Personally, I’ll be eager to see if the humor and flirting deficiencies of ChatGPT o1-preview have been solved in the final version.

You might also like…

• OpenAI Strawberry is here – it’s called o1-preview and it might be the most human ChatGPT ever
• ChatGPT Project Strawberry is incredibly smart and it could be available within the next two weeks
• If you think GPT-4o is something, wait until you see GPT-5 – a ‘significant leap forward’

Eric Hal Schwartz is a freelance writer for TechRadar with more than 15 years of experience covering the intersection of the world and technology. For the last five years, he served as head writer for Voicebot.ai and was on the leading edge of reporting on generative AI and large language models. He’s since become an expert on the products of generative AI models, such as OpenAI’s ChatGPT, Anthropic’s Claude, Google Gemini, and every other synthetic media tool. His experience runs the gamut of media, including print, digital, broadcast, and live events. Now, he’s continuing to tell the stories people want and need to hear about the rapidly evolving AI space and its impact on their lives. Eric is based in New York City.