# Deepseek Coder v2 vs Qwen2.5 72B Instruct

> Deepseek Coder v2 is the stronger model overall, scoring 35.9 to 31.9 on the Noometry Index.

- Canonical page: https://noometry.com/compare/deepseek-coder-v2-vs-qwen2-5-72b-instruct
- Last updated: 2026-10-11
- Shared benchmarks: 21

## Summary

- They share 21 benchmarks with published results for both. Deepseek Coder v2 scores higher in 4 categories and Qwen2.5 72B Instruct in 4 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in math, where Deepseek Coder v2 leads 34.9 to 19.3.

## Snapshot

| | Deepseek Coder v2 | Qwen2.5 72B Instruct |
|---|---|---|
| Provider | DeepSeek | Alibaba (Qwen) |
| Noometry Index | 35.9 | 31.9 |
| Rank | 220 | 267 |
| Context | — | 131K |
| Input $/M | — | $1.40 |
| Output $/M | — | $5.60 |
| Weights | Open | Open |

## Coding

- Deepseek Coder v2: 38.1 (#183)
- Qwen2.5 72B Instruct: 33.2 (#260)

| Benchmark | Deepseek Coder v2 | Qwen2.5 72B Instruct |
|---|---|---|
| BigCodeBench Instruct | 48.2% | 45.8% |
| LMArena Coding | 1251 | 1292 |
| BigCodeBench Complete | 59.7% | 55.9% |
| WeirdML | — | 16% |
| HumanEval+ | 82.3% | — |
| MBPP+ | 75.1% | — |

## Agentic & Tool Use

- Deepseek Coder v2: —
- Qwen2.5 72B Instruct: 22.1 (#133)

| Benchmark | Deepseek Coder v2 | Qwen2.5 72B Instruct |
|---|---|---|
| TheAgentCompany | — | 5.7% |
| BALROG | — | 16.2% |
| METR Time Horizons | — | 35.8% |

## Reasoning

- Deepseek Coder v2: 23.6 (#176)
- Qwen2.5 72B Instruct: 22.3 (#199)

| Benchmark | Deepseek Coder v2 | Qwen2.5 72B Instruct |
|---|---|---|
| LMArena Hard Prompts | 1207 | 1271 |
| WinoGrande | 83.7% | 82.3% |
| DTBench | — | 62.9% |
| LMCA | — | 13.4% |
| BIG-Bench Hard | — | 79.8% |
| Epoch Capabilities Index | — | 129 |
| ForecastBench | — | 57.5 |
| HellaSwag | — | 84.8% |
| PIQA | — | 82.6% |

## Math

- Deepseek Coder v2: 34.9 (#190)
- Qwen2.5 72B Instruct: 19.3 (#287)

| Benchmark | Deepseek Coder v2 | Qwen2.5 72B Instruct |
|---|---|---|
| LMArena Math | 1241 | 1283 |
| OTIS Mock AIME 2024-2025 | — | 8.1% |
| Omni-MATH | — | 33% |
| MATH Level 5 | — | 63.2% |
| GSM8K | 94.5% | — |

## Knowledge

- Deepseek Coder v2: 32.3 (#212)
- Qwen2.5 72B Instruct: 27.0 (#253)

| Benchmark | Deepseek Coder v2 | Qwen2.5 72B Instruct |
|---|---|---|
| LMArena Expert | 1181 | 1245 |
| ARC (AI2) Challenge | 64.3% | 94.5% |
| GPQA Diamond | — | 49.1% |
| MMLU-Pro | — | 63.1% |
| Confabulations | — | 19.1% |
| GPQA (HELM) | — | 42.6% |
| MMLU | — | 85.3% |
| TriviaQA | — | 71.9% |

## Multilingual

- Deepseek Coder v2: 36.3 (#240)
- Qwen2.5 72B Instruct: 41.0 (#213)

| Benchmark | Deepseek Coder v2 | Qwen2.5 72B Instruct |
|---|---|---|
| LMArena Non-English | 1182 | 1252 |
| LMArena Chinese | 1201 | 1272 |
| LMArena French | 1185 | 1280 |
| LMArena German | 1164 | 1234 |
| LMArena Japanese | 1126 | 1180 |
| LMArena Korean | 1104 | 1188 |
| LMArena Russian | 1188 | 1264 |
| LMArena Spanish | 1153 | 1256 |

## Instruction Following

- Deepseek Coder v2: 61.7 (#242)
- Qwen2.5 72B Instruct: 65.5 (#221)

| Benchmark | Deepseek Coder v2 | Qwen2.5 72B Instruct |
|---|---|---|
| LMArena Instruction Following | 1180 | 1254 |
| IFEval | — | 80.6% |

## Long Context

- Deepseek Coder v2: 37.0 (#224)
- Qwen2.5 72B Instruct: 38.9 (#188)

| Benchmark | Deepseek Coder v2 | Qwen2.5 72B Instruct |
|---|---|---|
| LMArena Longer Query | 1219 | 1282 |

## Writing & Preference

- Deepseek Coder v2: 38.2 (#253)
- Qwen2.5 72B Instruct: 46.7 (#215)

| Benchmark | Deepseek Coder v2 | Qwen2.5 72B Instruct |
|---|---|---|
| LMArena Text | 1191 | 1269 |
| LMArena Creative Writing | 1120 | 1221 |
| LMArena Multi-Turn | 1177 | 1272 |
| WildBench | — | 80.2% |

## FAQ

### Is Deepseek Coder v2 better than Qwen2.5 72B Instruct?

Deepseek Coder v2 is the stronger model overall, scoring 35.9 to 31.9 on the Noometry Index.

### Is Deepseek Coder v2 or Qwen2.5 72B Instruct better for coding?

Deepseek Coder v2 scores higher on coding benchmarks: 38.1 versus 33.2 in the Noometry coding category.

### How many benchmarks do Deepseek Coder v2 and Qwen2.5 72B Instruct share?

21 benchmarks have published results for both models. Deepseek Coder v2 has 24 scored results on Noometry and Qwen2.5 72B Instruct has 43.