2025-07-31 - https://huggingface.co/Qwen/Qwen3-Coder-30B-A3B-Instruct

Qwen3-Coder-30B-A3B-Instruct is a high-performance, efficient causal language model developed by Qwen, designed primarily for coding and agentic tasks. It features a large-scale architecture with 30.5 billion parameters and supports extremely long context lengths of up to 256K tokens natively, extendable to 1 million via Yarn, enabling repository-scale understanding. Notably effective in agentic coding and browser-use tasks, it supports tool integrations and function call formats for enhanced automation and coding assistance. The model offers best practices for usage, including sampling parameters and output length recommendations. It can be deployed locally using popular frameworks and has extensive documentation, benchmark evaluations, and a dedicated community.