AI BriefingThinking with Images
Key point
OpenAI unveiled o3 and o4-mini, which perform visual reasoning by directly cropping, zooming, and rotating images.
Details
OpenAI's latest o-series models, o3 and o4-mini, go beyond simple image recognition to demonstrate visual reasoning capabilities that directly utilize images within the chain-of-thought process.
These models can crop, zoom, and rotate images using image processing tools without the need for separate specialized models. This forms a new axis of test-time compute scaling that combines text and visual reasoning, achieving outstanding performance on multimodal benchmarks.
Key features include:
- Enhanced visual intelligence: The models automatically manipulate images to analyze them, accurately extracting information even from imperfect photos.
- Multimodal agent: Combined with Python data analysis, web search, and image generation capabilities, it provides an agentic experience for solving complex problems.
- Flexible interaction: Enables higher-order tasks such as rotating an image to read the contents of a notebook with upside-down handwriting, or finding and visualizing a path through a maze.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.
Related
Privatemode Benchmarks GLM-5.3-Flash as Vision-Capable Decision Model with Parity to Jev but Higher CostOpenAI Discloses AI Agents Accessed US Government Websites During Training