Qwen3-Coder series welcomes a new member - Qwen3-Coder-Flash, a programming model affectionately called "pastry-level" by developers. With its outstanding performance and efficient operation speed, it brings new surprises to the field of programming.
The full name of Qwen3-Coder-Flash is Qwen3-Coder-30B-A3B-Instruct. It performs excellently in terms of performance and efficiency, especially in areas such as agentic coding, agentic browser use, and tool use, demonstrating strong agent capabilities. This capability not only surpasses current top open-source models, but is slightly behind the top-tier version Qwen3-Coder-480B-A35B-Instruct, as well as leading closed-source models like Claude Sonnet-4 and GPT4.1, providing developers with powerful programming support.
In terms of context understanding, Qwen3-Coder-Flash also performs excellently. It natively supports 256K tokens and can be extended up to 1M tokens through YaRN, which means the entire project library can be understood, and developers no longer need to worry about issues with code context fragmentation. This feature is particularly important for handling large projects and complex codebases, significantly improving development efficiency.
Qwen3-Coder-Flash not only has powerful performance, but also good compatibility and ease of use. It supports multi-platform usage, and has a specially designed function call format, optimized for platforms such as Qwen Code, CLINE, Roo Code, and Kilo Code, making it more convenient for developers to use. Whether deploying on local hardware or using in the cloud, Qwen3-Coder-Flash can easily cope with it, providing great convenience for developers.
Currently, Qwen3-Coder-Flash has been officially open-sourced on the ModelScope community and Hugging Face, allowing developers to freely deploy it on local hardware and build their own code assistants. At the same time, the model is also available on Qwen Chat, where developers can go to chat.qwen.ai directly experience its powerful programming capabilities.
QwenChat: chat.qwen.ai
ModelScope Community: https://www.modelscope.cn/models/Qwen/Qwen3-Coder-30B-A3B-Instruct
Hugging Face: https://huggingface.co/Qwen/Qwen3-Coder-30B-A3B-Instruct