Qwen3-Coder-30B-A3B-Instruct (Coder-Flash)
Small 30B-A3B (3B active) Apache 2.0 coder with 256K context (1M via YaRN), aimed at local agentic coding in Qwen Code and Cline.
A streamlined sibling of the 480B Qwen3-Coder with a custom function-call format and repository-scale context. The hosted alias qwen3-coder-flash appeared in Alibaba Cloud Model Studio on 2025-07-29.
- Date
- Thursday, 31 July 2025
- Lab
- Alibaba (Qwen)
- Kind
- open-weights
- Access
- open weights
Figures
| Measure | Value | Measured by |
|---|---|---|
| Context length | 256K native, 1M with YaRN 30.5B total parameters per Hugging Face metadata | company |
Date is the Hugging Face repo creation (2025-07-31); the API alias predates it by two days. No benchmark table was opened.
Sources
- huggingface.co/Qwen/Qwen3-Coder-30B-A3B-Instruct
- www.alibabacloud.com/help/en/model-studio/newly-released-models
This record was checked against its sources on 6 October 2026. How we check