×
OpenAI releases o3-mini, an AI reasoning model great for science, math and coding
Written by
Published on
Join our daily newsletter for breaking news, product launches and deals, research breakdowns, and other industry-leading AI coverage
Join Now

OpenAI has released o3-mini, a new free STEM-focused reasoning model, in direct response to competitive pressure from Chinese AI company DeepSeek.

The big picture: OpenAI’s latest model release represents a significant shift in its strategy by making advanced reasoning capabilities freely available to all users for the first time.

  • The model shows particular strength in science, math, and coding applications while operating with lower costs and latency than its predecessor
  • Users can select from three different reasoning effort levels to balance speed and accuracy
  • The medium version of o3-mini delivers responses 24% faster than o1-mini, reducing average response time from 10.16 to 7.7 seconds

Performance improvements: The new model demonstrates substantial gains in both accuracy and user satisfaction compared to previous versions.

  • Testing revealed a 39% reduction in major errors compared to o1-mini
  • Users preferred o3-mini’s responses 56% of the time over its predecessor
  • The model includes a prototype search function that provides up-to-date answers with web source links

Accessibility and usage limits: OpenAI has implemented a tiered access system that balances availability with usage restrictions.

  • Paid subscribers (Plus, Team, and Pro tiers) will see o3-mini replace o1-mini in their model options
  • Plus and Team subscribers can send 150 messages per day, up from the previous 50-message limit
  • Free users can access the model through a “Reason” option in the ChatGPT interface

Safety and limitations: OpenAI has conducted extensive testing while acknowledging the model’s current constraints.

  • Safety evaluations included testing for potential misuse in areas like chemical and biological weapons
  • The model’s persuasion capabilities were found to be comparable to human-written text
  • Testing revealed poor performance in machine learning research capabilities and scored 0% on automating OpenAI research engineer tasks
  • Training utilized both public data and custom datasets with rigorous filtering for quality and risk mitigation

Strategic implications: This release signals OpenAI’s commitment to maintaining market leadership while democratizing access to AI capabilities, though questions remain about the long-term impact of making advanced AI models freely available and how competitors will respond to this move.

OpenAI hits back at DeepSeek with o3-mini reasoning model

Recent News

AI chip tracking bill aims to curb China’s tech access

New legislation would mandate location-tracking systems on export-controlled AI semiconductors to prevent smuggling to China, as bipartisan concerns grow over technology diversion that could enhance military capabilities.

AI transforms motivation and learning approaches, study finds

AI systems showcase learning without traditional motivation, suggesting educators should focus more on designing effective learning environments rather than inspiring students.

Chrome’s new AI mode blocks scams: How it safeguards users

Chrome's new AI protection system analyzes suspicious patterns to identify and block previously undetected scams across Google's platforms, keeping users twice as safe from phishing attempts.