The Qwen2.5-Coder-32B model has achieved the top ranking on both Hugging Face Models and Gradio Spaces, highlighting its popularity in the AI community. The Qwen team has also released a comprehensive set of evaluation and fine-tuning tools, along with demos for the Qwen2.5-Coder series, which can be run locally with Gradio's artifacts mode. Additionally, the CodeQwen1.5-7B model has been trained for code synthesis to tackle puzzles for the ARC Prize competition, demonstrating its capabilities despite compute constraints. The latest Qwen model is now integrated into the Fragments by E2B app, which is free and open-source, allowing users to run generated code in E2B sandboxes. Furthermore, the Qwen-2.5-Coder and New Sonnet 3.5 models are now available on Copilot Arena, inviting users to participate in building a leaderboard.