Buried the lead. Microsoft fine tuned Qwen2.5-VL-7B. That’s the big conversation starter here. Have any of the big providers done this before?

“The model is based on Qwen2.5-VL-7B and trained with supervised fine-tuning.”