Qwen 3 Coder 480B

Qwen3 Coder 480B A35B is a large-scale Mixture-of-Experts model with 480 billion total parameters and 35 billion active per forward pass. Its 262K native context window is extendable to 1M with Yarn, enabling processing of entire codebases. It sets state-of-the-art results among open models across agentic coding, browser-use, and tool-use benchmarks, with performance reported as comparable to leading proprietary systems such as Claude Sonnet 4 on coding tasks. Trained on a large suite of verifiable coding tasks with executable environments to enable reinforcement learning from real code execution, the model is open-weight under the Apache 2.0 license with broad inference-framework support.