Vision

Apache 2.0 (7B); Qwen Research License, non-commercial (3B); Qwen License (72B)Open weights where releasedUpdated August 2026

Qwen2.5 VL

Qwen2.5 VL is a vision-language Qwen model family for image, screenshot, and document workflows.

Alibaba Qwen · Qwen

Editorial review

Reviewed byOpenSourcesAI EditorialLast updatedAugust 2026SourcesCanonical artifact, Representative model card

Model checkpoints, context windows, provider support, local runtime compatibility, and license terms can change quickly. Verify the exact model card before production or commercial use.

Best for

Vision-language assistants, document images, and multimodal agent workflows.

Who should use it

  • Vision-language assistants, document images, and multimodal agent workflows.
  • Builders who want local or self-hosted testing options.

Common workflows

  • Vision-language and document image workflows
  • vision workflows
  • multimodal workflows
  • documents workflows

Deployment and hardware notes

The representative 7B Instruct checkpoint needs ~5.0 GB at Q4_K_M (8.3B parameters) — comfortable on an 8 GB card once image preprocessing is accounted for.

License and usage notes

Apache 2.0 (7B); Qwen Research License, non-commercial (3B); Qwen License (72B). Open weights where released. Verify the exact model card and license terms for the checkpoint or hosted provider you use.

Strengths

  • Open weights where released model option for Qwen workflows.
  • Vision-language assistants, document images, and multimodal agent workflows.
  • The 12.8M-pixel ceiling is a processor setting, not a hard limit of the model: lower it to trade visual detail for context budget and speed.

Limitations

  • Licensing is not uniform across the family — the 7B is Apache 2.0, the 3B is a non-commercial research licence and the 72B carries the Qwen licence — so the size you pick determines what you may ship. Superseded by Qwen3-VL for new work.
  • The representative 7B Instruct checkpoint needs ~5.0 GB at Q4_K_M (8.3B parameters) — comfortable on an 8 GB card once image preprocessing is accounted for.
  • Context window and limits: 128,000 tokens.
  • Verify the exact model card, provider docs, license, and serving support before production use.

Local workflow notes

The 12.8M-pixel ceiling is a processor setting, not a hard limit of the model: lower it to trade visual detail for context budget and speed.

Local runtimes: Ollama where supported, LM Studio where supported, llama.cpp where supported, Transformers

Platforms: Windows, macOS, Linux

Vision spec

This page covers a family of checkpoints. The figures below describe Qwen2.5-VL-7B-Instruct, its representative release — other sizes in the family differ.

MemoryVaries by size; the 7B Instruct checkpoint needs ~5.0 GB at Q4_K_M (8.3B parameters)Image inputDynamic resolution, 3,136 to 12.8M pixels, 14-pixel patches; video inputContext128,000 tokens

The generation that made Qwen-VL a default for document and agent work: it reads charts and forms, points at things by returning coordinates, and takes video natively through a temporal patch size of 2.

Sources to verify

Canonical artifactRepresentative model card

Related resources

Continue with model source notes, local tools, and implementation guides related to this model.

Qwen family hubAI tools directoryModel selection guideModel source notesModels directory
HardwareVaries by size; the 7B Instruct checkpoint needs ~5.0 GB at Q4_K_M (8.3B parameters)RuntimeOllama or LM Studio where supported, llama.cpp, Transformers, vLLMContext128,000 tokensLast updated2026
Canonical artifact →

Model ecosystem connections

Use these next-step links to move from this profile into related tools, comparisons, guides, stacks, and curated shortlists.

Recommended runtimes and tools

Ollama →LM Studio →llama.cpp →Transformers →vLLM →

Setup and deployment

Check if this model fits your GPU →What is quantization? GGUF, Q4, Q8 explained →What is VRAM and how much do you need? →

Related model pages

Qwen family hub →Qwen2 VL →Qwen3 14B →Qwen3 235B A22B →Qwen3 VL →Qwen2.5 32B →Qwen2.5 72B →

Guides, stacks, and comparisons

AI tools directory →Model selection guide →All model comparisons →

Related models

Qwen2 VLVision-language baseline comparisons and multimodal prototypes.Qwen3 14BMultilingual chat, assistant workflows, and Qwen-family comparisons.Qwen3 235B A22BBuilders testing frontier-style open-weight reasoning and coding in hosted or multi-GPU environments.Qwen3 VLBuilders adding visual understanding to open AI workflows.Qwen2.5 32B32B open-weight model in the Qwen family.Qwen2.5 72B72B open-weight model in the Qwen family.

Ready to run this model locally?

Find a compatible interface in our Local AI Tools directory →