Models

← All models

Qwen2.5-VL-3B-Instruct

Alibaba · 2025-01-26 · 3.8B parameters

Qwen2.5-VL-3B-Instruct is Alibaba's compact instruction-tuned 3B-class checkpoint in the Qwen2.5-VL family. It accepts text, image, and video inputs and generates text, with dynamic-resolution vision, OCR and document understanding, visual grounding, long-video understanding, and visual-agent capabilities. Its published configuration sets 128,000 maximum positions, while the model-card guidance identifies 32,768 tokens as the original text context and prescribes YaRN for longer text inputs.

Benchmark scores