Qwen3-Coder: Alibaba claims world’s most advanced agentic AI model for coding

Chinese tech giant Alibaba Group has announced the launch of open-source AI coding model, Qwen3-Coder, claiming it is the most advanced agentic AI coding model to date. Alibaba said Qwen3-Coder is designed for high-performance software development, and excels in agentic AI coding tasks — from generating new code and managing complex coding workflows to debugging across entire codebases.

Designed to help with coding and using tools online, the newly open-sourced AI model has a total of 480 billion parameters but uses only 35 billion parameters per token, making it efficient without sacrificing performance. The efficiency is significant as the model uses fewer resources but still claims to have achieved "competitive results against leading state-of-the-art (SOTA) models" across key benchmarks in agentic coding, browser use, and tool use.

Qwen3-Coder enters the fast-evolving field of advanced AI coding assistants, competing with both proprietary models like OpenAI's GPT-4o with Code Interpreter, Anthropic's Claude 3.5 Sonnet, and Google’s Gemini 1.5 Pro, as well as open-source rivals such as DeepSeek-Coder V2 and Meta’s Code Llama 70B, all vying for a larger share in developer workflows.

Additionally, Alibaba said it is open-sourcing Qwen Code, a powerful command-line tool that lets developers talk to AI in natural language to get coding tasks done. It is designed to work smoothly with the Qwen3-Coder AI model, which it does by optimising custom prompts and interaction protocols. The model also works well with tools like the Claude Code interface, making it even easier for developers to execute their coding tasks.

Qwen3-Coder is trained on a large mix of code and text and can handle large inputs. "Trained on an extensive dataset of code and general text data, Qwen3-Coder is engineered for robust agentic coding. It natively supports a context window of 256K tokens, extendable up to 1 million tokens, enabling it to process vast codebases in a single session. Its superior performance stems not only from scaling across tokens, context length, and synthetic data during pre-training, but also from innovative post-training techniques such as long-horizon reinforcement learning (agent RL)," the company said in a statement.

The upgrade lets Qwen3-Coder tackle complex, real-world coding problems via multi-step interactions with external tools. "Qwen3-Coder achieves SOTA performance among open-source models on SWE-Bench Verified (a benchmark for evaluating AI models' ability to solve real-world software issues), even without test-time or inference scaling."

Notably, agentic AI coding is transforming software development by enabling more autonomous, efficient, and accessible programming workflows. With its open-source availability, strong agentic coding capabilities, and seamless compatibility with popular developer tools and interfaces, Alibaba aims to position Qwen3-Coder as a valuable tool for global software developers.

The model is available on Hugging Face and GitHub, while developers can also access the model on Qwen Chat or via cost-effective APIs through Model Studio, Alibaba's generative AI development platform.

Alibaba claimed Qwen-based coding models have already surpassed 20 million downloads globally. Earlier this year, it unveiled AI models like Qwen 2.5 Max and Qwen 3, which claimed to outperform DeepSeek’s R1 model and even the likes of OpenAI’s GPT-4o and Meta’s Llama.

The Qwen3-Coder launch comes amid AI major OpenAI's recent launch of ChatGPT Agent, a new tool designed to handle complex, multi-step tasks using its own virtual computer.