Main to remember
- The O3 model of Openai may think with images and use tools such as web navigation and python to solve complex problems and in several steps.
- O4-minini is optimized for speed and profitability, outperforming previous models in mathematical, coding and non-stem tasks.
Openai has published two new IA models today, O3 and O4-Mini, expanding its capacities of reasoning and visual intelligence.
O3 represents the most advanced reasoning model of the company to date, while O4-MINI offers improved performance for mathematical, coding and visual tasks at lower costs.
O3 is the first model of the O series capable of using all the available Chatppt tools, including web navigation, python, image understanding and generation. The two models introduce “reflection with images”, allowing direct integration of visual entries into their reasoning process.
The O3 model establishes new references in software engineering, mathematics and scientific reasoning, going beyond O1 in tasks requiring a detailed analysis, a generation of hypotheses and an interpretation of the visual content. External tests show that the O3 reduces major errors by 20% compared to the O1.
O4-minini, optimized for high-speed performance, ranks first in references, notably likes 2024 and 2025, demonstrating high precision between the rod and non-semester fields.
OPENAI has also published Codex CLI, a local coding agent for the execution of models from the terminal. A $ 1 million grant program will help developers build with it.
The two models have undergone security tests using the OPENAI updated preparation framework, assessments confirming the levels of risk lower than the thresholds of biosecurity, cybersecurity and self-improvement categories.
The models are available today for Chatgpt Plus, Pro and Team users, replacing O1 and O3-Mini. Chatgpt Enterprise and EDU customers will have access next week. Free level users can test O4-Mini using the “Think” option before requests.
OPENAI plans to launch O3-Pro in the coming weeks, combining O3’s capabilities with full management of tools for advanced reasoning tasks.
16,643
Lunam Points