# DeepSeek V4 Pro vs Qwen2.5 72B Instruct

> DeepSeek V4 Pro is the stronger model overall, scoring 54.2 to 31.5 on the Noometry Index.

- Canonical page: https://noometry.com/compare/deepseek-v4-pro-vs-qwen2-5-72b-instruct
- Last updated: 2026-10-10
- Shared benchmarks: 24

## Summary

- They share 24 benchmarks with published results for both. DeepSeek V4 Pro scores higher in 9 categories and Qwen2.5 72B Instruct in 0 categories; 9 gaps are clear of the uncertainty.
- The widest gap is in math, where DeepSeek V4 Pro leads 64.6 to 18.1.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 98.6% for DeepSeek V4 Pro and 8.1% for Qwen2.5 72B Instruct.
- DeepSeek V4 Pro is cheaper at $0.43 / $0.87 per million input/output tokens, against $1.40 / $5.60 for Qwen2.5 72B Instruct.
- DeepSeek V4 Pro accepts more context: 1M tokens versus 131K.

## Snapshot

| | DeepSeek V4 Pro | Qwen2.5 72B Instruct |
|---|---|---|
| Provider | DeepSeek | Alibaba (Qwen) |
| Noometry Index | 54.2 | 31.5 |
| Rank | 28 | 259 |
| Context | 1M | 131K |
| Input $/M | $0.43 | $1.40 |
| Output $/M | $0.87 | $5.60 |
| Weights | Open | Open |

## Coding

- DeepSeek V4 Pro: 52.2 (#31)
- Qwen2.5 72B Instruct: 32.8 (#256)

| Benchmark | DeepSeek V4 Pro | Qwen2.5 72B Instruct |
|---|---|---|
| WeirdML | 66.2% | 16% |
| LMArena Coding | 1470 | 1293 |
| SWE-bench Verified | 77.6% | — |
| FrontierCode | 28.6% | — |
| LMArena WebDev | 1582 | — |
| SciCode | 51% | — |
| BigCodeBench Instruct | — | 45.8% |
| BigCodeBench Complete | — | 55.9% |
| ALE-Bench | 1,403 | — |

## Agentic & Tool Use

- DeepSeek V4 Pro: 32.7 (#54)
- Qwen2.5 72B Instruct: 22.0 (#124)

| Benchmark | DeepSeek V4 Pro | Qwen2.5 72B Instruct |
|---|---|---|
| APEX-Agents | 47.3% | — |
| TheAgentCompany | — | 5.7% |
| BALROG | — | 16.2% |
| METR Time Horizons | — | 35.8% |
| Vending-Bench 2 | 3,285 | — |

## Reasoning

- DeepSeek V4 Pro: 56.4 (#22)
- Qwen2.5 72B Instruct: 21.4 (#197)

| Benchmark | DeepSeek V4 Pro | Qwen2.5 72B Instruct |
|---|---|---|
| LMArena Hard Prompts | 1461 | 1271 |
| DTBench | 93.9% | 62.9% |
| LMCA | 45.5% | 13.4% |
| Epoch Capabilities Index | 155.39 | 129.01 |
| ForecastBench | 56.1 | 57.5 |
| ARC-AGI-2 | 61.3% | — |
| Kagi LLM Benchmark | 53.5% | — |
| NYT Connections (extended) | 91.3% | — |
| ARC-AGI-1 | 90.5% | — |
| CritPt | 18% | — |
| Chess Puzzles | 47% | — |
| Mystery Game Puzzles | 43% | — |
| Surface Evolver Bench | 40% | — |
| BIG-Bench Hard | — | 79.8% |
| HellaSwag | — | 84.8% |
| PIQA | — | 82.6% |
| WinoGrande | — | 82.3% |

## Math

- DeepSeek V4 Pro: 64.6 (#24)
- Qwen2.5 72B Instruct: 18.1 (#281)

| Benchmark | DeepSeek V4 Pro | Qwen2.5 72B Instruct |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 98.6% | 8.1% |
| LMArena Math | 1455 | 1283 |
| FrontierMath (Tiers 1-3) | 64.6% | — |
| FrontierMath Tier 4 | 26.8% | — |
| MathArena Final-Answer Competitions | 76.6% | — |
| ProofBench | 50% | — |
| Omni-MATH | — | 33% |
| MATH Level 5 | — | 63.2% |

## Knowledge

- DeepSeek V4 Pro: 59.5 (#28)
- Qwen2.5 72B Instruct: 26.9 (#247)

| Benchmark | DeepSeek V4 Pro | Qwen2.5 72B Instruct |
|---|---|---|
| GPQA Diamond | 91.7% | 49.1% |
| LMArena Expert | 1463 | 1245 |
| SimpleQA Verified | 52.9% | — |
| MMLU-Pro | — | 63.1% |
| Confabulations | — | 19.1% |
| Vectara Hallucination Rate | 8.6% | — |
| GPQA (HELM) | — | 42.6% |
| ARC (AI2) Challenge | — | 94.5% |
| MMLU | — | 85.3% |
| TriviaQA | — | 71.9% |

## Multilingual

- DeepSeek V4 Pro: 54.5 (#44)
- Qwen2.5 72B Instruct: 41.2 (#208)

| Benchmark | DeepSeek V4 Pro | Qwen2.5 72B Instruct |
|---|---|---|
| LMArena Non-English | 1438 | 1252 |
| LMArena Chinese | 1486 | 1272 |
| LMArena French | 1473 | 1279 |
| LMArena German | 1458 | 1234 |
| LMArena Japanese | 1448 | 1180 |
| LMArena Korean | 1452 | 1188 |
| LMArena Russian | 1452 | 1264 |
| LMArena Spanish | 1457 | 1255 |

## Instruction Following

- DeepSeek V4 Pro: 76.3 (#44)
- Qwen2.5 72B Instruct: 65.7 (#216)

| Benchmark | DeepSeek V4 Pro | Qwen2.5 72B Instruct |
|---|---|---|
| LMArena Instruction Following | 1448 | 1254 |
| IFEval | — | 80.6% |

## Long Context

- DeepSeek V4 Pro: 45.0 (#49)
- Qwen2.5 72B Instruct: 38.9 (#183)

| Benchmark | DeepSeek V4 Pro | Qwen2.5 72B Instruct |
|---|---|---|
| LMArena Longer Query | 1459 | 1282 |
| CL-bench Life | 13.5% | — |

## Writing & Preference

- DeepSeek V4 Pro: 65.6 (#45)
- Qwen2.5 72B Instruct: 46.9 (#210)

| Benchmark | DeepSeek V4 Pro | Qwen2.5 72B Instruct |
|---|---|---|
| LMArena Text | 1451 | 1269 |
| LMArena Creative Writing | 1446 | 1221 |
| LMArena Multi-Turn | 1467 | 1272 |
| EQ-Bench Creative Writing | 1553 | — |
| WildBench | — | 80.2% |
| EQ-Bench 4 | 1166 | — |

## FAQ

### Is DeepSeek V4 Pro better than Qwen2.5 72B Instruct?

DeepSeek V4 Pro is the stronger model overall, scoring 54.2 to 31.5 on the Noometry Index.

### Which is cheaper, DeepSeek V4 Pro or Qwen2.5 72B Instruct?

DeepSeek V4 Pro is cheaper. It lists at $0.43 per million input tokens and $0.87 per million output tokens; Qwen2.5 72B Instruct lists at $1.40 and $5.60.

### Is DeepSeek V4 Pro or Qwen2.5 72B Instruct better for coding?

DeepSeek V4 Pro scores higher on coding benchmarks: 52.2 versus 32.8 in the Noometry coding category.

### Which has the bigger context window?

DeepSeek V4 Pro does, with 1M tokens against 131K.

### How many benchmarks do DeepSeek V4 Pro and Qwen2.5 72B Instruct share?

24 benchmarks have published results for both models. DeepSeek V4 Pro has 48 scored results on Noometry and Qwen2.5 72B Instruct has 43.