OpenAI has released a new AI model, known as “o1”, aimed at addressing more complex reasoning tasks in science, coding, and mathematics. It’s designed to spend more time processing tasks before delivering responses; an approach that OpenAI claims improves their ability to solve difficult problems and “represents a new level of AI capability”.
Subscribers to ChatGPT Plus and Team can now test out two models in the series:
- o1-preview – which has stronger reasoning and broader knowledge, and can tackle “complex problems in science, coding, maths, and related fields”
- o1-mini – which is smaller, cheaper and designed for coding tasks
OpenAI plans to roll out access to Enterprise and Edu subscribers soon, and to bring o1-mini to all users for free in the future.
So far, the preview release lacks many of ChatGPT’s features, such as web browsing and the ability to handle multimedia inputs. Access is also currently limited to 50 queries per week for o1-preview and 50 queries per day for the o1-mini version.
OpenAI’s statement regarding the new release emphasises the importance of safety in the development of these models. These developments come amid increasing scrutiny of AI platforms’ data practices, as demonstrated by the EU’s investigation into Google’s AI, which raises questions about how personal data is processed during the training of AI models. As these new AI systems evolve, concerns around privacy remain central to discussions about their broader application and regulation.