AlibabaQwen3 Coder
Qwen3-Coder-480B-A35B-Instruct
The large instruction-tuned Qwen3 coding release.
Practical use
Practical guidance. Use repository tasks that require finding the relevant files before editing. Track whether a proposed fix addresses the original requirement and preserves nearby behaviour, rather than judging only whether the generated code looks plausible.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Access | Downloadable |
| Model identifiers | Qwen/Qwen3-Coder-480B-A35B-Instruct |
| Architecture | 480 billion total and 35 billion active parameters; 62 layers. Non-thinking output.Qwen3-Coder-480B-A35B-Instruct model card |
| Context | 262,144 native tokens; the card documents extension to about one million.Qwen3-Coder-480B-A35B-Instruct model card |
| Licence | Apache 2.0.Qwen3-Coder-480B-A35B-Instruct model card |
Related models
- Qwen3-Coder-Next
Alibaba · Qwen3 Coder
A published coding-agent model with a hybrid architecture.
Launch date not confirmedDownloadable - Qwen3-Coder-30B-A3B-Instruct
Alibaba · Qwen3 Coder
A smaller instruction-tuned checkpoint for coding agents.
Launch date not confirmedDownloadable