Qwen3‑Coder: Open‑Source AI Programming Agent That Beats the Competition
Alibaba’s Tongyi team unveiled the open‑source Qwen3‑Coder, a massive 450‑billion‑parameter programming model that outperforms leading closed‑source solutions, supports up to 1 M token context, offers a free CLI tool, and demonstrates impressive code generation capabilities across animations, games, and real‑world tasks.
Qwen3‑Coder Open‑Source Programming Agent
Alibaba Tongyi’s new open‑source model Qwen3‑Coder has set a new AI programming SOTA, outperforming DeepSeek V3, Kimi K2 and even the closed‑source Claude Sonnet 4.
The model family includes several sizes; the top‑tier Qwen3‑Coder‑480B‑A35B‑Instruct is a 450‑billion‑parameter MoE model with 35 B active parameters.
It natively supports a 256 K context window and, with YaRN, can be extended to 1 M tokens.
A command‑line tool called Qwen Code , built on Gemini Code, is also released as open source.
Demo prompts show that a single sentence can generate colorful p5.js animations, a 3‑D interactive globe, dynamic weather cards, mini‑games, a résumé template, and even a Minesweeper game.
Technical Details
During pre‑training, Qwen3‑Coder scales data, context, and synthetic data, using 7.5 T tokens of which 70 % are code, improving programming ability while preserving general and mathematical skills.
In the fine‑tuning stage, the team applies Scaling Code RL to generate diverse test cases and Scaling Long‑Horizon RL on Alibaba Cloud, running 20 000 parallel environments, achieving open‑source SOTA on SWE‑bench Verified.
No more $200 monthly subscription for Claude Code.
The model is released under the Apache License 2.0, making it commercial‑friendly and freely available.
Getting Started
Try it directly on the Qwen website.
Install the Qwen Code CLI and use the OpenAI SDK.
Apply for an API on Alibaba Cloud Bailei platform to combine with tools like Claude Code or Cline.
Project repository: https://github.com/QwenLM/qwen-code
Official site: https://chat.qwen.ai/
DataFunTalk
Dedicated to sharing and discussing big data and AI technology applications, aiming to empower a million data scientists. Regularly hosts live tech talks and curates articles on big data, recommendation/search algorithms, advertising algorithms, NLP, intelligent risk control, autonomous driving, and machine learning/deep learning.
How this landed with the community
Was this worth your time?
0 Comments
Thoughtful readers leave field notes, pushback, and hard-won operational detail here.