# DeepSeek V4 Pro vs Qwen2-72B

> DeepSeek V4 Pro is the stronger model overall, scoring 54.2 to 29.5 on the Noometry Index.

- Canonical page: https://noometry.com/compare/deepseek-v4-pro-vs-qwen2-72b
- Last updated: 2026-10-10
- Shared benchmarks: 20

## Summary

- They share 20 benchmarks with published results for both. DeepSeek V4 Pro scores higher in 9 categories and Qwen2-72B in 0 categories; 9 gaps are clear of the uncertainty.
- The widest gap is in knowledge, where DeepSeek V4 Pro leads 59.5 to 21.2.
- The biggest single-benchmark swing is WeirdML: 66.2% for DeepSeek V4 Pro and 11.3% for Qwen2-72B.

## Snapshot

| | DeepSeek V4 Pro | Qwen2-72B |
|---|---|---|
| Provider | DeepSeek | Alibaba (Qwen) |
| Noometry Index | 54.2 | 29.5 |
| Rank | 28 | 294 |
| Context | 1M | — |
| Input $/M | $0.43 | — |
| Output $/M | $0.87 | — |
| Weights | Open | Open |

## Coding

- DeepSeek V4 Pro: 52.2 (#31)
- Qwen2-72B: 28.7 (#303)

| Benchmark | DeepSeek V4 Pro | Qwen2-72B |
|---|---|---|
| WeirdML | 66.2% | 11.3% |
| LMArena Coding | 1470 | 1196 |
| SWE-bench Verified | 77.6% | — |
| FrontierCode | 28.6% | — |
| LMArena WebDev | 1582 | — |
| SciCode | 51% | — |
| BigCodeBench Instruct | — | 38.5% |
| BigCodeBench Complete | — | 54% |
| ALE-Bench | 1,403 | — |

## Agentic & Tool Use

- DeepSeek V4 Pro: 32.7 (#54)
- Qwen2-72B: 16.9 (#138)

| Benchmark | DeepSeek V4 Pro | Qwen2-72B |
|---|---|---|
| APEX-Agents | 47.3% | — |
| TheAgentCompany | — | 1.1% |
| METR Time Horizons | — | 29.9% |
| Vending-Bench 2 | 3,285 | — |

## Reasoning

- DeepSeek V4 Pro: 56.4 (#22)
- Qwen2-72B: 22.2 (#184)

| Benchmark | DeepSeek V4 Pro | Qwen2-72B |
|---|---|---|
| LMArena Hard Prompts | 1461 | 1191 |
| Epoch Capabilities Index | 155.39 | 125.27 |
| 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% | — |
| DTBench | 93.9% | — |
| LMCA | 45.5% | — |
| Surface Evolver Bench | 40% | — |
| ForecastBench | 56.1 | — |

## Math

- DeepSeek V4 Pro: 64.6 (#24)
- Qwen2-72B: 27.9 (#232)

| Benchmark | DeepSeek V4 Pro | Qwen2-72B |
|---|---|---|
| LMArena Math | 1455 | 1235 |
| FrontierMath (Tiers 1-3) | 64.6% | — |
| FrontierMath Tier 4 | 26.8% | — |
| MathArena Final-Answer Competitions | 76.6% | — |
| OTIS Mock AIME 2024-2025 | 98.6% | — |
| ProofBench | 50% | — |
| MATH Level 5 | — | 39.1% |

## Knowledge

- DeepSeek V4 Pro: 59.5 (#28)
- Qwen2-72B: 21.2 (#270)

| Benchmark | DeepSeek V4 Pro | Qwen2-72B |
|---|---|---|
| GPQA Diamond | 91.7% | 40.8% |
| LMArena Expert | 1463 | 1171 |
| SimpleQA Verified | 52.9% | — |
| Vectara Hallucination Rate | 8.6% | — |
| MMLU | — | 82.4% |

## Multilingual

- DeepSeek V4 Pro: 54.5 (#44)
- Qwen2-72B: 36.0 (#239)

| Benchmark | DeepSeek V4 Pro | Qwen2-72B |
|---|---|---|
| LMArena Non-English | 1438 | 1176 |
| LMArena Chinese | 1486 | 1240 |
| LMArena French | 1473 | 1169 |
| LMArena German | 1458 | 1151 |
| LMArena Japanese | 1448 | 1111 |
| LMArena Korean | 1452 | 1083 |
| LMArena Russian | 1452 | 1169 |
| LMArena Spanish | 1457 | 1168 |

## Instruction Following

- DeepSeek V4 Pro: 76.3 (#44)
- Qwen2-72B: 62.0 (#236)

| Benchmark | DeepSeek V4 Pro | Qwen2-72B |
|---|---|---|
| LMArena Instruction Following | 1448 | 1181 |

## Long Context

- DeepSeek V4 Pro: 45.0 (#49)
- Qwen2-72B: 36.1 (#230)

| Benchmark | DeepSeek V4 Pro | Qwen2-72B |
|---|---|---|
| LMArena Longer Query | 1459 | 1193 |
| CL-bench Life | 13.5% | — |

## Writing & Preference

- DeepSeek V4 Pro: 65.6 (#45)
- Qwen2-72B: 41.0 (#236)

| Benchmark | DeepSeek V4 Pro | Qwen2-72B |
|---|---|---|
| LMArena Text | 1451 | 1203 |
| LMArena Creative Writing | 1446 | 1181 |
| LMArena Multi-Turn | 1467 | 1195 |
| EQ-Bench Creative Writing | 1553 | — |
| EQ-Bench 4 | 1166 | — |

## FAQ

### Is DeepSeek V4 Pro better than Qwen2-72B?

DeepSeek V4 Pro is the stronger model overall, scoring 54.2 to 29.5 on the Noometry Index.

### Is DeepSeek V4 Pro or Qwen2-72B better for coding?

DeepSeek V4 Pro scores higher on coding benchmarks: 52.2 versus 28.7 in the Noometry coding category.

### How many benchmarks do DeepSeek V4 Pro and Qwen2-72B share?

20 benchmarks have published results for both models. DeepSeek V4 Pro has 48 scored results on Noometry and Qwen2-72B has 26.