Qwen3-VL-8B-Instruct model trained for the AgenticOCR project: query-driven, on-demand visual evidence extraction for document RAG.

This repository contains the model weights, configuration, tokenizer, chat template, and visual preprocessor files required for inference.