What's New :
14th September 2024 (8 Topics)

OpenAI’s new AI models o1 and o1-mini (Project Strawberry)

Context

OpenAI has launched new AI models named o1 and o1-mini. The o1 model is part of OpenAI’s secretive Project Strawberry, focusing on enhancing AI's reasoning abilities. It is designed to tackle complex tasks in science, coding, and mathematics, the o1 model represents a significant leap in AI capabilities.

About

  • The o1 model is engineered to approach problems more thoughtfully, mimicking human-like problem-solving. It evaluates queries from multiple perspectives, checks its outputs, and learns from errors.
  • Although the new models lack the ability to browse the web or process uploaded files and images like the GPT-4o model, they can refine their “thinking process” by trying new strategies and recognising mistakes.
  • Performance:
    • In a mathematics contest, the o1 model solved 83% of problems, a major improvement over earlier versions that solved only 13%.
    • In coding tasks, it outperformed 89% of participants.
  • Availability and Access:
    • Current Access: Available in ChatGPT for Plus and Team subscribers, with a weekly message limit (30 for o1-preview and 50 for o1-mini).
    • Future Updates: OpenAI plans to expand availability and introduce a feature for automatic model selection based on user prompts.

Differences Between o1 and o1-mini

  • o1 Model: It offers advanced reasoning and problem-solving abilities. It is suitable for complex tasks and applications requiring deep problem-solving capabilities.
  • o1-mini Model: It is smaller, faster, and more cost-effective version of the o1 model. It is 80% cheaper than the o1 model. It is designed for efficiency in coding tasks and applications where extensive world knowledge is less critical.

Verifying, please be patient.

Enquire Now