Alibaba Cloud / Qwen Team model product
Qwen2.5-Coder is a specialized coding model trained on 5.5 trillion tokens of code data, supporting 92 programming languages with a 128K context window. It excels in code generation, completion, repair, and multi-programming tasks while maintaining strong performance in mathematics and general capabilities.
Updated Aug 10, 2026. Default version: Qwen2.5-Coder 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-Coder 32B Instruct.
| BigCodeBench-Full | 0.5 | 1 | 1 | 100.0% | C | |
| BigCodeBench-Hard | 0.3 | 1 | 1 | 100.0% | C | |
| MBPP | 0.9 | 3 | 37 | 94.4% | C | |
| TheoremQA | 0.4 | 3 | 6 | 60.0% | C | |
| HumanEval | 0.9 | 7 | 76 | 92.0% | C | |
| Winogrande | 0.8 | 10 | 22 | 57.1% | C | |
| TruthfulQA | 0.5 | 14 | 18 | 23.5% | C | |
| HellaSwag | 0.8 | 16 | 27 | 42.3% | C | |
| ARC-C | 0.7 | 21 | 34 | 39.4% | C | |
| GSM8k | 0.9 | 24 | 48 | 51.1% | C | |
| MMLU-Redux | 0.8 | 42 | 48 | 12.8% | C | |
| MATH | 0.6 | 51 | 71 | 28.6% | C | |
| LiveCodeBench | 0.3 | 60 | 73 | 18.1% | C | |
| MMLU | 0.8 | 77 | 100 | 23.2% | C | |
| MMLU-Pro | 0.5 | 115 | 129 | 10.9% | C |
Preference and agent-evaluation signals from published Arena datasets.
| text | coding | 245 | 1276.1 | 873 | N/A | |
| text | expert | 246 | 1219.8 | 267 | N/A | |
| text | hard prompts english | 248 | 1271.2 | 759 | N/A | |
| text style control | coding | 250 | 1342.3 | 873 | N/A | |
| text style control | expert | 250 | 1275.2 | 267 | N/A | |
| text | math | 251 | 1250.6 | 725 | N/A | |
| text | industry mathematical | 253 | 1243.2 | 619 | N/A | |
| text | hard prompts | 254 | 1251.6 | 1,386 | N/A | |
| text | industry legal and government | 255 | 1256.8 | 273 | N/A | |
| text style control | math | 255 | 1269.8 | 725 | N/A | |
| text style control | industry legal and government | 256 | 1309.2 | 273 | N/A | |
| text | longer query | 257 | 1251.1 | 804 | N/A | |
| text style control | hard prompts english | 257 | 1319.3 | 759 | N/A | |
| text | industry business and management and financial operations | 258 | 1223.6 | 629 | N/A | |
| text | russian | 258 | 1227.3 | 822 | N/A | |
| text style control | russian | 258 | 1270.1 | 822 | N/A | |
| text | chinese | 259 | 1220.9 | 528 | N/A | |
| text | industry software and it services | 259 | 1262.0 | 1,381 | N/A | |
| text style control | hard prompts | 260 | 1303.1 | 1,386 | N/A | |
| text style control | industry business and management and financial operations | 261 | 1283.0 | 629 | N/A | |
| text style control | industry mathematical | 261 | 1268.9 | 619 | N/A | |
| text style control | industry software and it services | 263 | 1319.6 | 1,381 | N/A | |
| text | instruction following | 264 | 1223.2 | 2,227 | N/A | |
| text style control | chinese | 264 | 1279.9 | 528 | N/A | |
| text style control | longer query | 268 | 1286.8 | 804 | N/A | |
| text | overall | 270 | 1230.2 | 5,432 | N/A | |
| text | exclude ties | 270 | 1126.7 | 3,370 | N/A | |
| text | multi turn | 270 | 1222.3 | 1,024 | N/A | |
| text | english | 271 | 1248.1 | 2,652 | N/A | |
| text style control | instruction following | 271 | 1264.0 | 2,227 | N/A | |
| text | industry medicine and healthcare | 274 | 1185.7 | 221 | N/A | |
| text | non english | 276 | 1204.3 | 2,780 | N/A | |
| text style control | multi turn | 277 | 1250.9 | 1,024 | N/A | |
| text style control | non english | 278 | 1247.0 | 2,780 | N/A | |
| text style control | exclude ties | 279 | 1184.4 | 3,370 | N/A | |
| text | industry life and physical and social science | 280 | 1221.8 | 956 | N/A | |
| text style control | overall | 280 | 1270.5 | 5,432 | N/A | |
| text style control | english | 280 | 1286.3 | 2,652 | N/A | |
| text style control | industry medicine and healthcare | 282 | 1254.5 | 221 | N/A | |
| text style control | industry life and physical and social science | 285 | 1277.4 | 956 | N/A | |
| text | creative writing | 289 | 1174.2 | 836 | N/A | |
| text | industry writing and literature and language | 290 | 1191.0 | 1,540 | N/A | |
| text style control | industry writing and literature and language | 290 | 1227.5 | 1,540 | N/A | |
| text style control | creative writing | 292 | 1206.8 | 836 | N/A | |
| text | industry entertainment and sports and media | 294 | 1150.7 | 929 | N/A | |
| text style control | industry entertainment and sports and media | 303 | 1187.4 | 929 | N/A |
Official vendor API PAYG pricing is summarized first. The table then lists individual provider offerings without treating their minimum as the official price.
| Nvidia | qwen/qwen2.5-coder-32b-instruct | global | N/A | N/A | 128K | |
| Alibaba (China) | qwen2-5-coder-32b-instruct | global | $0.287 | $0.861 | 131.1K | |
| Cloudflare Workers AI | @cf/qwen/qwen2.5-coder-32b-instruct | global | $0.66 | $1 | 32.8K | |
| Kilo Gateway | qwen/qwen-2.5-coder-32b-instruct | global | $0.66 | $1 | 32.8K | |
| OpenRouter | qwen/qwen-2.5-coder-32b-instruct | global | $0.66 | $1 | 32.8K | |
| Cloudflare AI Gateway | workers-ai/@cf/qwen/qwen2.5-coder-32b-instruct | global | $0.66 | $1 | 128K |
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-Coder 32B Instruct | 45.4 | 32B | 128K | 128K | No | Apache 2.0 |
A concise description based on the published model registry.
Qwen2.5-Coder is a specialized coding model trained on 5.5 trillion tokens of code data, supporting 92 programming languages with a 128K context window. It excels in code generation, completion, repair, and multi-programming tasks while maintaining strong performance in mathematics and general capabilities.
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 Coder 32B.
Qwen2.5 Coder 32B's default version was released on Sep 19, 2024.
Qwen2.5 Coder 32B's official API price is $0.287 per million input tokens and $0.861 per million output tokens via Alibaba (China). The lowest tracked third-party offer starts at $0.66 input and $1 output via Cloudflare Workers AI.
Qwen2.5 Coder 32B is published under Alibaba Cloud / Qwen Team in the model registry.
The default version has a 128K token context window.
No. The default version is not marked as having publicly available weights.
6 published provider offerings are linked to the default version.
Nearby ranked alternatives include Phi 4 Reasoning, Gemma 2 27B, Sarvam 30B.