Qwen 2.5 Coder 32B

Alibaba's state-of-the-art open-source code model achieving GPT-4o-level coding performance. Released November 2024 with 32B parameters - small enough to run on consumer hardware (32GB+ RAM). Best open-source model on multiple code generation benchmarks (EvalPlus, LiveCodeBench, BigCodeBench). Scores 73.7 on Aider code repair (comparable to GPT-4o). Excels across 40+ programming languages with 65.9 on McEval multi-language benchmark. Achieves 75.2 on MdEval code repair (first among open-source). Available in sizes from 0.5B to 32B. Represents breakthrough in accessible, GPT-4o-class coding assistance.