Alibaba Cloud / Qwen Team model product
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.
Updated Aug 10, 2026. Default version: Qwen2.5 32B Instruct
Structured fields from the published default version.
This profile uses the latest version under this unique model that has a calculated LLMBoard score. Arena and price are excluded.
Published benchmark records for the scored version Qwen2.5 32B Instruct.
| MMLU-STEM | 0.8 | 1 | 2 | 100.0% | C | |
| TheoremQA | 0.4 | 2 | 6 | 80.0% | C | |
| BBH | 0.8 | 4 | 12 | 72.7% | C | |
| MultiPL-E | 0.8 | 6 | 13 | 58.3% | C | |
| MBPP | 0.8 | 7 | 37 | 83.3% | C | |
| Winogrande | 0.8 | 8 | 22 | 66.7% | C | |
| GSM8k | 1.0 | 9 | 48 | 83.0% | C | |
| TruthfulQA | 0.6 | 11 | 18 | 41.2% | C | |
| HellaSwag | 0.9 | 13 | 27 | 53.9% | C | |
| MATH | 0.8 | 15 | 71 | 80.0% | C | |
| ARC-C | 0.7 | 22 | 34 | 36.4% | C | |
| HumanEval | 0.9 | 25 | 76 | 68.0% | C | |
| MBPP+ | 0.7 | 26 | 37 | 30.6% | C | |
| MMLU-Redux | 0.8 | 35 | 48 | 27.7% | C | |
| MMLU | 0.8 | 48 | 100 | 52.5% | C | |
| HumanEval+ | 0.5 | 70 | 76 | 8.0% | C | |
| MMLU-Pro | 0.7 | 86 | 129 | 33.6% | C | |
| GPQA | 0.5 | 178 | 233 | 23.7% | C |
Preference and agent-evaluation signals from published Arena datasets.
The default version has no published Arena rows, or its source alias has not been resolved.
Official vendor API PAYG pricing is summarized first. The table then lists individual provider offerings without treating their minimum as the official price.
| Alibaba (China) | qwen2-5-32b-instruct | global | $0.287 | $0.861 | 131.1K | |
| Alibaba | qwen2-5-32b-instruct | global | $0.70 | $2.8 | 131.1K |
Official prices use only the vendor's configured official Provider and positive standard USD PAYG rates. Third-party offers remain explicitly labeled.
All published versions linked to this unique model. The score columns identify the version used by the current overall ranking.
| Qwen2.5 32B Instruct | 56.9 | 32.5B | N/A | N/A | No | Apache 2.0 |
A concise description based on the published model registry.
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.
Use the benchmark, Arena and pricing sections above as separate evidence. A missing field means the current data snapshot does not support that claim.
Data snapshot: 2026-08-07. Editorial model content is not available in the backend.
Open a comparison with the three ranked models immediately above and below this model.
Recommendations prioritize the same model type and family, then the closest published LLMBoard score.
Common questions about Qwen2.5 32B.
Qwen2.5 32B's default version was released on Sep 19, 2024.
Qwen2.5 32B's official API price is $0.70 per million input tokens and $2.8 per million output tokens via Alibaba.
Qwen2.5 32B is published under Alibaba Cloud / Qwen Team in the model registry.
The current registry does not publish a context window for the default version.
No. The default version is not marked as having publicly available weights.
2 published provider offerings are linked to the default version.
Nearby ranked alternatives include GPT-4.5, Qwen3 VL 235B A22B Thinking, Qwen3.5 27B.