The biggest change: o3 and o4-mini can now think using images as part of their reasoning process. Uploaded visuals can also be handled even if blurry or rotated — with the models able to adjust them using its own tools.

→ View original post on X — @rowancheung