OpenAI’s o1 model has leaked online before its official full release. The leak occurred on Friday and allowed users to access the full version of the model for about two hours. The new model features native image capabilities, allowing it to analyze images and provide detailed step-by-step explanations.
The o1 model is a significant improvement over previous models, including GPT-4o, which powers Advanced Voice. CEO Sam Altman has confirmed that o1 will be a divergence from the usual GPT-style models released by OpenAI.
According to sources, the leak was due to an issue with limited external access being prepared for the model. The problem has now been fixed. Despite this, the leak has provided insight into the full capabilities of the o1 model, including its ability to reason over images and solve problems.
One example of the model’s potential was seen in a user who got it to solve an image puzzle, spending nearly two minutes thinking through the problem. The AI also demonstrated its ability to walk through every single element of an image showing a recent SpaceX rocket launch.
The full version of o1 is expected to be released sometime this year, although no official date has been confirmed. However, most AI companies are holding back until after the US Presidential election.
OpenAI’s o1 model will be able to analyze images, access web search tools, and perform data analysis. The leak has sparked excitement among tech enthusiasts, who see this as a significant advancement in AI technology.
Source: https://www.tomsguide.com/ai/chatgpt/openais-o1-model-leaked-on-friday-and-it-is-wild-heres-what-happened