OpenAI's o3 model reduces errors by 20% with visual reasoning

OpenAI has unveiled two new AI models, o3 and o4-mini, marking significant advancements in reasoning and visual intelligence. The o3 model is the company's most advanced reasoning model to date, while o4-mini offers enhanced performance for mathematical, coding, and visual tasks at a lower cost.

One of the standout features of the o3 model is its ability to independently utilize all available ChatGPT tools, including web browsing, Python, image understanding, and generation. This capability allows o3 to integrate visual inputs directly into its reasoning process, a feature referred to as “thinking with images.” This innovation represents a significant leap forward in the field of AI, enabling more comprehensive and nuanced problem-solving.

The o3 model has set new benchmarks in software engineering, mathematics, and scientific reasoning, outperforming its predecessor, o1, in tasks that require detailed analysis, hypothesis generation, and visual content interpretation. External testing has shown that o3 reduces major errors by 20% compared to o1, highlighting its superior accuracy and reliability.

Ask Aime: What impact do the new AI models from OpenAI have on the AI market?

On the other hand, o4-mini is optimized for high-throughput performance and has ranked first in benchmarks such as AIME 2024 and 2025. This model demonstrates strong accuracy across both stem and non-STEM fields, making it a versatile tool for a wide range of applications.

In addition to these new models, OpenAI has also released Codex CLI, a local coding agent designed to run models from the terminal. This tool is expected to streamline the development process for developers, providing them with a more efficient way to integrate AI models into their projects. To support this initiative, OpenAI has announced a $1 million grant program aimed at encouraging developers to build with Codex CLI.

These advancements by OpenAI underscore the company's commitment to pushing the boundaries of AI technology. The introduction of o3 and o4-mini, along with Codex CLI, represents a significant step forward in the field of AI, offering new capabilities and efficiencies that are poised to have a profound impact on various industries.