OpenMed-SynthVision-MedVL-AIO-GGUF

OpenMed's MedVL series—comprising Qwen2.5-3B-MedVL, Qwen3.5-2B-MedVL, and Ministral-3B-MedVL—are lightweight, specialized medical vision-language models fine-tuned from their respective strong open-source bases (Qwen2.5-VL-3B, Qwen3.5-2B, Ministral-3B) for clinical applications like radiology report generation, medical VQA, pathology slide analysis, dermatology lesion identification, and multimodal diagnostics across X-rays, CT/MRI scans, histopathology images, and ophthalmic fundus photos. These ~2-3B parameter models prioritize edge deployment on laptops/single GPUs via efficient architectures (Qwen's dynamic resolution ViT + Gated DeltaNet, Ministral's optimized SLM design), achieving high-fidelity medical reasoning, anatomical localization, disease classification, and report structuring while preserving layout/spatial awareness for real-world hospital workflows.

Model Files

Qwen2.5-3B-MedVL

Qwen3.5-2B-MedVL

Ministral-3B-MedVL

Model Sources