# Claude 3.7 Sonnet vs DeepSeek V4 Pro

> DeepSeek V4 Pro is the stronger model overall, scoring 54.2 to 39.4 on the Noometry Index.

- Canonical page: https://noometry.com/compare/claude-3-7-sonnet-vs-deepseek-v4-pro
- Last updated: 2026-10-10
- Shared benchmarks: 25

## Summary

- They share 25 benchmarks with published results for both. Claude 3.7 Sonnet scores higher in 2 categories and DeepSeek V4 Pro in 7 categories; 9 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where DeepSeek V4 Pro leads 56.4 to 18.7.
- The biggest single-benchmark swing is ARC-AGI-1: 28.6% for Claude 3.7 Sonnet and 90.5% for DeepSeek V4 Pro.
- DeepSeek V4 Pro has downloadable open weights; the other is API-only.

## Snapshot

| | Claude 3.7 Sonnet | DeepSeek V4 Pro |
|---|---|---|
| Provider | Anthropic | DeepSeek |
| Noometry Index | 39.4 | 54.2 |
| Rank | 154 | 28 |
| Context | — | 1M |
| Input $/M | — | $0.43 |
| Output $/M | — | $0.87 |
| Weights | Proprietary | Open |

## Coding

- Claude 3.7 Sonnet: 41.1 (#120)
- DeepSeek V4 Pro: 52.2 (#31)

| Benchmark | Claude 3.7 Sonnet | DeepSeek V4 Pro |
|---|---|---|
| SWE-bench Verified | 61% | 77.6% |
| LMArena Coding | 1361 | 1470 |
| FrontierCode | — | 28.6% |
| SWE-bench Verified (bash only) | 52.8% | — |
| Aider Polyglot | 64.9% | — |
| LMArena WebDev | — | 1582 |
| SciCode | — | 51% |
| GSO | 3.8% | — |
| WeirdML | — | 66.2% |
| LiveBench Coding | 74.5% | — |
| CadEval | 54% | — |
| ALE-Bench | — | 1,403 |

## Agentic & Tool Use

- Claude 3.7 Sonnet: 33.9 (#46)
- DeepSeek V4 Pro: 32.7 (#54)

| Benchmark | Claude 3.7 Sonnet | DeepSeek V4 Pro |
|---|---|---|
| APEX-Agents | — | 47.3% |
| TheAgentCompany | 30.9% | — |
| Cybench | 20% | — |
| DeepResearch Bench | 43.6% | — |
| OSWorld | 35.8% | — |
| METR Time Horizons | 60% | — |
| Vending-Bench 2 | — | 3,285 |

## Reasoning

- Claude 3.7 Sonnet: 18.7 (#252)
- DeepSeek V4 Pro: 56.4 (#22)

| Benchmark | Claude 3.7 Sonnet | DeepSeek V4 Pro |
|---|---|---|
| ARC-AGI-2 | 0.9% | 61.3% |
| ARC-AGI-1 | 28.6% | 90.5% |
| LMArena Hard Prompts | 1333 | 1461 |
| Epoch Capabilities Index | 141.16 | 155.39 |
| ForecastBench | 61.8 | 56.1 |
| SimpleBench | 46.4% | — |
| Kagi LLM Benchmark | — | 53.5% |
| NYT Connections (extended) | — | 91.3% |
| CritPt | — | 18% |
| Chess Puzzles | — | 47% |
| EnigmaEval | 4.2% | — |
| LiveBench Reasoning | 87.8% | — |
| Mystery Game Puzzles | — | 43% |
| DTBench | — | 93.9% |
| LiveBench Data Analysis | 74% | — |
| LMCA | — | 45.5% |
| Surface Evolver Bench | — | 40% |
| LiveBench | 76.1% | — |

## Math

- Claude 3.7 Sonnet: 35.7 (#137)
- DeepSeek V4 Pro: 64.6 (#24)

| Benchmark | Claude 3.7 Sonnet | DeepSeek V4 Pro |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 57.8% | 98.6% |
| LMArena Math | 1336 | 1455 |
| FrontierMath (Tiers 1-3) | — | 64.6% |
| FrontierMath Tier 4 | — | 26.8% |
| MathArena Final-Answer Competitions | — | 76.6% |
| ProofBench | — | 50% |
| Omni-MATH | 33% | — |
| LiveBench Math | 79% | — |
| MATH Level 5 | 91.2% | — |
| FrontierMath (Feb 2025 set) | 4.1% | — |

## Knowledge

- Claude 3.7 Sonnet: 39.7 (#117)
- DeepSeek V4 Pro: 59.5 (#28)

| Benchmark | Claude 3.7 Sonnet | DeepSeek V4 Pro |
|---|---|---|
| GPQA Diamond | 79.7% | 91.7% |
| LMArena Expert | 1321 | 1463 |
| Humanity's Last Exam | 8% | — |
| SimpleQA Verified | — | 52.9% |
| MMLU-Pro | 78.4% | — |
| Confabulations | 14.7% | — |
| Vectara Hallucination Rate | — | 8.6% |
| GPQA (HELM) | 60.8% | — |

## Multimodal

- Claude 3.7 Sonnet: 33.1 (#90)
- DeepSeek V4 Pro: —

| Benchmark | Claude 3.7 Sonnet | DeepSeek V4 Pro |
|---|---|---|
| LMArena Vision | 1169 | — |
| GeoBench | 68% | — |
| VPCT | 39% | — |
| SpatialViz-Bench | 33.9% | — |

## Multilingual

- Claude 3.7 Sonnet: 44.3 (#174)
- DeepSeek V4 Pro: 54.5 (#44)

| Benchmark | Claude 3.7 Sonnet | DeepSeek V4 Pro |
|---|---|---|
| LMArena Non-English | 1296 | 1438 |
| LMArena Chinese | 1299 | 1486 |
| LMArena French | 1302 | 1473 |
| LMArena German | 1301 | 1458 |
| LMArena Japanese | 1268 | 1448 |
| LMArena Korean | 1249 | 1452 |
| LMArena Russian | 1311 | 1452 |
| LMArena Spanish | 1299 | 1457 |

## Instruction Following

- Claude 3.7 Sonnet: 73.1 (#120)
- DeepSeek V4 Pro: 76.3 (#44)

| Benchmark | Claude 3.7 Sonnet | DeepSeek V4 Pro |
|---|---|---|
| LMArena Instruction Following | 1352 | 1448 |
| LiveBench Instruction Following | 81.3% | — |
| IFEval | 83.4% | — |

## Long Context

- Claude 3.7 Sonnet: 50.2 (#10)
- DeepSeek V4 Pro: 45.0 (#49)

| Benchmark | Claude 3.7 Sonnet | DeepSeek V4 Pro |
|---|---|---|
| LMArena Longer Query | 1373 | 1459 |
| Fiction.LiveBench | 83.3% | — |
| CL-bench Life | — | 13.5% |

## Writing & Preference

- Claude 3.7 Sonnet: 54.6 (#145)
- DeepSeek V4 Pro: 65.6 (#45)

| Benchmark | Claude 3.7 Sonnet | DeepSeek V4 Pro |
|---|---|---|
| LMArena Text | 1314 | 1451 |
| LMArena Creative Writing | 1332 | 1446 |
| EQ-Bench Creative Writing | 1412 | 1553 |
| LMArena Multi-Turn | 1339 | 1467 |
| Short-Story Creative Writing | 81.1% | — |
| WildBench | 81.4% | — |
| EQ-Bench 4 | — | 1166 |
| LiveBench Language | 59.9% | — |

## FAQ

### Is Claude 3.7 Sonnet better than DeepSeek V4 Pro?

DeepSeek V4 Pro is the stronger model overall, scoring 54.2 to 39.4 on the Noometry Index.

### Is Claude 3.7 Sonnet or DeepSeek V4 Pro better for coding?

DeepSeek V4 Pro scores higher on coding benchmarks: 52.2 versus 41.1 in the Noometry coding category.

### How many benchmarks do Claude 3.7 Sonnet and DeepSeek V4 Pro share?

25 benchmarks have published results for both models. Claude 3.7 Sonnet has 58 scored results on Noometry and DeepSeek V4 Pro has 48.