Try Delta
← All models

Alibaba2025

Qwen3

Hybrid reasoning · dense & MoE

Think when you need it, answer fast when you don’t. Qwen3 switches thinking/non-thinking modes for coding, math, and 119 languages, from 0.6B on a phone to 30B-A3B MoE on a workstation.

  • Fits fromPhone / edge · 4 GB+
  • Contextup to 33K
  • Sizes6
  • RunsLocal on Delta
ReasoningCodingMultilingualMath & code119 languagesThink / fast modes

Sizes

Qwen3 0.6B

1.5 GB • 33k ctx • F16

Qwen3 1.7B

1.1 GB • 33k ctx • F16

Qwen3 4B

2.5 GB • 33k ctx • Q4_K_M

Qwen3 8B

5 GB • 33k ctx • Q4_K_M

Qwen3 14B

9 GB • 33k ctx • Q4_K_M

Qwen3 30B-A3B

20 GB • 33k ctx • Q4_K_M

Memory at context length

Approximate working set for Qwen3 1.7B as context grows.

4K
1.4 GB
8K
1.5 GB
16K
1.5 GB
33K
1.6 GB

Related families

Qwen 3.8

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

Qwen3.5

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

Qwen3 VL

See, read, and act on screen: images, video, OCR, and GUI agents on your machine.