Try Delta
← All models

Alibaba2025

Qwen3 VL

Vision-language · dense & MoE

See, read, and act on what’s on screen. Qwen3 VL handles images, video, OCR in 32 languages, GUI agents, and spatial reasoning, so Delta can work with photos, PDFs, and UI captures without sending them anywhere.

  • Fits fromPhone / edge · 5 GB+
  • Contextup to 8K
  • Sizes4
  • RunsLocal on Delta
VisionOCRAgentsMultilingualDocument OCRGUI agentsImage + text

Sizes

Qwen3 VL 2B

1.8 GB • 8k ctx • Q4_K_M

Qwen3 VL 4B

2.5 GB • 8k ctx • Q4_K_M

Qwen3 VL 8B

5.5 GB • 8k ctx • Q4_K_M

Qwen3 VL 30B-A3B

20 GB • 8k ctx • Q4_K_M

Memory at context length

Approximate working set for Qwen3 VL 2B as context grows.

4K
2.2 GB
8K
2.4 GB

Related families

Qwen3.5

A natively multimodal daily driver: text, vision, and compact reasoning on your machine.

Qwen 3.8

Qwen 3.8 Distill models: highly efficient and capable models optimized for edge and desktop.

Qwen3

Think when you need it, answer fast when you don’t: coding, math, and 119 languages, local.