Almost all of the latest models are MLLMs (multimodal large language models). For exmaple, [1] evaluates GPT-6 Astra on vision tasks.

[1] https://blog.roboflow.com/gpt-6-astra-vision/