Qwen2.5 Coder 32B Instruct
Qwen · released 2024-11-11 · apache-2.0 license
Qwen2.5-Coder-32B-Instruct was the strongest open code model at its Nov-2024 release (Apache-2.0): vendor LiveCodeBench 31.4 (2407-2409 window) and Aider 73.7, comparable to GPT-4o on code repair.
Key specs
| Type | Local open-weight |
|---|---|
| Parameters | 32.76B total |
| Architecture | qwen2 |
| Context window | 128K tokens |
| Knowledge cutoff | 2024-06-30 |
| Modalities | text |
| Recommended backends | — |
| Minimum viable rig | — |
Benchmark scores
| GPQA Diamond | — |
|---|---|
| SWE-bench Verified | — |
| AIME | — |
| MMLU-Pro | — |
| BFCL v3 (tool use) | — |
| Composite score | — |
| Community rating | No reviews yet |
VRAM & disk per quantization
| Quant | VRAM | Disk | RAM | Context |
|---|---|---|---|---|
| Q4_K_M | 20.5 GB | 19 GB | — | 128K |
Strengths & weaknesses
Strengths: SOTA open-source code model at release, matching GPT-4o on several code benchmarks; Strong code repair (Aider 73.7); Apache-2.0
Weaknesses: Coding-specialized; weaker general breadth; No native reasoning mode