OpenAI has launched two new AI reasoning models, o3 and o4-mini, which are designed to enhance ChatGPT's capabilities significantly. These models are trained to think longer before responding and can utilize all available tools within ChatGPT, including web searches, file analysis with Python, visual reasoning, and image generation. This advancement allows them to produce detailed answers efficiently, typically within a minute, making them adept at solving complex, multi-faceted problems.
Key Features of OpenAI o3
OpenAI o3 is the most powerful model in this release, excelling in areas such as coding, math, science, and visual perception. It sets new state-of-the-art (SOTA) benchmarks on various tests, including Codeforces and SWE-bench. Notably, o3 reduces major errors by 20% compared to its predecessor, o1, particularly in programming, business, and creative ideation. Early testers praised its analytical capabilities and its ability to generate and evaluate hypotheses in fields like biology and engineering.
Key Features of OpenAI o4-mini
OpenAI o4-mini is a smaller, cost-effective model that delivers impressive performance in math, coding, and visual tasks. It is recognized as the best-performing model on AIME 2024 and 2025 benchmarks and outperforms o3-mini in non-STEM tasks and data science. Its efficiency allows for higher usage limits, making it suitable for high-volume queries that require reasoning.
Improvements and Competitive Landscape
Both models have been rated by external experts as having better instruction-following capabilities and providing more useful, verifiable responses than previous versions. They also feature enhanced conversational abilities, referencing memory and past interactions for personalized responses. This release is part of OpenAI's strategy to maintain a competitive edge against other AI companies like Google, Meta, and Anthropic, as the field increasingly focuses on reasoning models to enhance performance.
