Try Delta
← All models

Qwen3.5

Alibaba · 2026 · Natively multimodal reasoning

Alibaba’s natively multimodal 3.5 line. Thinking and non-thinking modes for text and vision, from a sub-1B companion to a 9B daily driver. Strong default if you want one family for chat, images, and light agents.

Maker
Alibaba
Architecture
Natively multimodal reasoning
Fits from
Phone / edge · 4 GB+ RAM
Context
up to 33K
Runs
Local Intelligence
VisionReasoningMultilingualDaily driverText + visionCompact reasoning

Sizes

Qwen3.5 0.8BQ8_0

requires 4 GB+ memory · 33K context · Phone / edge

831 MBWeights
Qwen3.5 2BQ8_0

requires 6 GB+ memory · 33K context · Phone / edge

2 GBWeights
Qwen3.5 4BQ4_K_M

requires 7 GB+ memory · 33K context · Laptop

2.7 GBWeights
Qwen3.5 9BQ4_K_M

requires 11 GB+ memory · 33K context · Laptop

5.7 GBWeights

Memory at context length

Approximate working set for Qwen3.5 0.8B as context grows.

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

Related families

Qwen3 VL

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

Qwen3

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

Qwen3 Coder

A coding specialist for agentic engineering: function calling, tools, and repo-scale work locally.