OpenAI has launched its new AI model, o3-mini, a cost-efficient reasoning model designed to improve performance in tasks requiring advanced reasoning, such as mathematics, coding, and scientific inquiries. Released on January 31, 2025, this model is available through OpenAI’s ChatGPT platform and APIs, marking the first time a reasoning model is accessible to free-tier users. It also replaces the o1-mini model in OpenAI’s offerings.
Try search + reasoning together in ChatGPT.
— OpenAI (@OpenAI) January 31, 2025
Free users can use OpenAI o3-mini with search by selecting the Search + Reason buttons together. Paid users can select o3-mini from the model picker. https://t.co/8SGPBLFQ3P pic.twitter.com/S8bFBBG8yh
o3-mini family
The o3-mini model introduces several enhancements over its predecessor, o1-mini:
- Reasoning Effort Control: Users can adjust the cognitive load (low, medium, high) to balance speed and accuracy.
- Structured Outputs: Supports JSON Schema constraints for generating well-defined outputs.
- Developer Messages: Replaces system messages with a more structured “developer” role for better instruction handling.
- Function Calling and Tools Support: Enables integration with external tools for automation.
- Improved Speed and Accuracy: Offers 24% faster response times and reduces major errors by 39% compared to o1-mini.
- Deliberative Alignment: A new safety mechanism that enhances the model’s ability to evaluate prompts for compliance with safety standards.
The model is optimized for enterprise use, offering significant cost reductions—63% cheaper than o1-mini and 93% cheaper than the full o1 model—while maintaining robust reasoning capabilities. Pricing starts at $1.10 per million input tokens and $4.40 per million output tokens.
Availability and Access
o3-mini is available to:
- Free-tier ChatGPT users with limited access via the “Reason” option.
- Paid ChatGPT Plus, Team, and Pro users with increased rate limits (e.g., Pro users enjoy unlimited access).
- Developers through OpenAI’s Chat Completions API, Assistants API, and Batch API.
Enterprise access will begin in February 2025.
About OpenAI
OpenAI has positioned o3-mini as a response to increasing competition in the AI space, particularly from DeepSeek’s R1 model. While DeepSeek’s open-source approach has gained attention for its affordability, OpenAI aims to maintain its leadership with ChatGPT by offering a reasoning model that balances cost-efficiency with advanced capabilities. This launch follows OpenAI’s preview of the o3 series in December 2024.
