Qwen2.5 32B Instruct
Qwen2.5-32B-Instruct is an instruction-tuned 32 billion parameter language model, part of the Qwen2.5 series. It is designed to follow instructions, generate long texts (over 8K tokens), understand structured data (e.g., tables), and generate structured outputs, especially JSON. The model supports multilingual capabilities across over 29 languages.
Benchmark results
| Benchmark | Score | Tags | Source |
|---|---|---|---|
| ARC-C | 70.4% | self-reported llm-stats | link → |
| BBH | 84.5% | self-reported llm-stats | link → |
| GPQA | 49.5% | self-reported llm-stats | link → |
| GSM8k | 95.9% | self-reported llm-stats | link → |
| HellaSwag | 85.2% | self-reported llm-stats | link → |
| HumanEval | 88.4% | self-reported llm-stats | link → |
| HumanEval+ | 52.4% | self-reported llm-stats | link → |
| MATH | 83.1% | self-reported llm-stats | link → |
| MBPP | 84.0% | self-reported llm-stats | link → |
| MBPP+ | 67.2% | self-reported llm-stats | link → |
| MMLU | 83.3% | self-reported llm-stats | link → |
| MMLU-Pro | 69.0% | self-reported llm-stats | link → |
| MMLU-Redux | 83.9% | self-reported llm-stats | link → |
| MMLU-STEM | 80.9% | self-reported llm-stats | link → |
| MultiPL-E | 75.4% | self-reported llm-stats | link → |
| TheoremQA | 44.1% | self-reported llm-stats | link → |
| TruthfulQA | 57.8% | self-reported llm-stats | link → |
| Winogrande | 82.0% | self-reported llm-stats | link → |