Qwen2.5-VL-7B-Instruct

Qwen2.5-VL-7B-Instruct is a 7-billion-parameter instruction-tuned vision-language model from Alibaba's Qwen team that understands images, documents, charts, UI layouts, and video. It supports visual question answering, OCR-based document parsing, object grounding, and structured JSON/coordinate outputs for agentic visual tasks. Source: supabase.