A sensible choice for developers running a coding agent locally, with only 3 billion active parameters and free weights.
Kenning's take
What it does
Qwen3-Coder-Next is an open-weight coding model built for agents and local development, released in early 2026. It has 80 billion parameters with only 3 billion active, a 256K context window and a hybrid attention design. Qwen says it works with tools such as Claude Code, Qwen Code, Qoder, Cline and Trae.
- 80B total, 3B active parameters
- 262,144-token native context
- Hybrid Gated DeltaNet and attention layout
- Non-thinking mode only
- Apache 2.0 open weights, FP8 and GGUF versions
Who it suits. Developers running coding agents locally or cheaply.
Key facts
- Category
- AI models
- Part of
- Qwen
- Made by
- Alibaba (Qwen team)
- Based in
- China
- Released
- 30 January 2026
- How to get it
- Free download from Hugging Face
- Works on
- self-hosted, Hugging Face
- In the UK
- Yes
- Free plan
- Yes, open weights
FAQ
Is Qwen3-Coder-Next free?
Yes. The weights are open under Apache 2.0.
How many parameters does Qwen3-Coder-Next have?
80 billion in total, with 3 billion active.
Which tools work with Qwen3-Coder-Next?
Qwen lists Claude Code, Qwen Code, Qoder, Kilo, Trae and Cline.
Facts from the maker's own pages and the sources listed. Checked 2026-10-11.
Similar tools
MAI-UI
Alibaba Tongyi research models that operate smartphone apps by looking at the screen and tapping
Qwen-Image-2.0
Alibaba's 7-billion-parameter model that both creates and edits images, with strong text rendering at 2K
Qwen-Image-Layered
Open Alibaba model that splits an image into separate transparent layers so each part can be edited alone
Qwen-Image-3.0
Alibaba's hosted image model built for text-heavy layouts, taking prompts up to 4,500 tokens