Qwen2.5-Coder-32B is an LLM that can code well that runs on my Mac

Simon Willison reviews the Qwen2.5-Coder Series from Alibaba’s research team. The 32B model version is open-source and capable of running on a MacBook Pro M2, which Willison tested. His assessments based on both Qwen’s scores and Paul Gauthier’s Aider benchmarks suggest that Qwen2.5-Coder-32B is highly competitive, offering comparison with existing models such as GPT-4o and Claude 3.5. He also provides detailed insights into running different versions of the model on his Mac.

• Qwen2.5-Coder-32B is claimed to match GPT-4o coding capabilities.
• The 32B model is Apache 2.0 licensed and open-source.
• Benchmark scores are impressive, showing competitive results.
• Runs on a 64GB MacBook Pro M2 without quitting other apps.
• Installation and execution of the model are explained by Willison.