OpenAI is releasing a new, smaller, and more efficient artificial intelligence model called o3-mini for free, targeting the excitement generated by the recently unveiled open-source model R1 from the Chinese AI startup DeepSeek. Scheduled for release on January 31, o3-mini boasts advanced reasoning capabilities that can effectively deconstruct complex problems to find solutions. OpenAI announced that o3-mini will be available to all ChatGPT Plus, Team, and Pro users, while free-tier users will have limited access. The model aims to push the boundaries of what smaller models can accomplish. To prepare this model, OpenAI has been employing PhD students for research and development efforts, as detailed in a recent job posting that highlights the goal of creating challenging coding tests for large language models. DeepSeek’s R1 model has disrupted the US tech industry by being available for free, pressuring companies like Google and Anthropic to adjust their pricing strategies.

OpenAI is keen to reinforce its leading position in AI development and commercialization, particularly considering that R1 may have utilized outputs from OpenAI’s models in its training process. While o3-mini may not compete directly with R1 on price, it emphasizes improved efficiency and excels in math, science, and coding tasks. The model will also feature capabilities like web search integration, function calls from user code, and varying levels of reasoning to adjust speed and problem-solving abilities. Additionally, DeepSeek’s quick ascent raises concerns regarding US efforts to limit China’s AI advancements. Previous US administrations have imposed sanctions to restrict China’s access to advanced Nvidia chips used in building top-tier AI models, although it remains unclear which specific chips DeepSeek utilized in its development.