Has OpenAI basically delivered AGI with o3 and o4 mini?

OpenAI just dropped two new AI models that have the tech world buzzing with a provocative question: did they just casually release artificial general intelligence (AGI) without making a big deal about it?

Some heavy hitters in AI are making bold claims. One OpenAI model trainer admitted being “tempted to call this model AGI,” while economist Tyler Cowen went further, declaring “I think it is AGI, honestly.” Even Sam Altman seemed to validate these sentiments by quoting a tweet claiming the new models show “genius level” capabilities.

What are these new models?

OpenAI released two advanced models that push AI reasoning to new frontiers:

• o3 – Their most powerful reasoning model yet, showing breakthrough capabilities in coding, math, science, and visual perception
• o4 mini – A smaller, cost-efficient model that still achieves remarkable performance, particularly in handling complex math problems

Thinking with images: a game-changing feature

The most impressive innovation might be what OpenAI calls “thinking with images.” Unlike previous AI systems that just passively process images, these new models actually think with images as part of their reasoning process. This means:

• The AI can actively zoom in on portions of images
• It can rotate, crop, and manipulate images to better understand them
• It can deblur and enhance low-quality images
• It integrates images directly into its reasoning chains
• It can cross-reference visual information with web knowledge

In practice, users report remarkable capabilities – from solving complex math problems written on upside-down sticky notes in a pile of toys to accurately reading tiny handwritten text by zooming and enhancing images.