# Claude Opus 4 vs DeepSeek-V3

> Claude Opus 4 is the stronger model overall, scoring 43.1 to 39.5 on the Noometry Index. DeepSeek-V3 costs 74× less per token, which makes it the better buy when Claude Opus 4's lead doesn't matter for your workload.

- Canonical page: https://noometry.com/compare/claude-opus-4-vs-deepseek-v3
- Last updated: 2026-10-11
- Shared benchmarks: 41

## Summary

- They share 41 benchmarks with published results for both. Claude Opus 4 scores higher in 8 categories and DeepSeek-V3 in 0 categories; 7 gaps are clear of the uncertainty.
- The widest gap is in math, where Claude Opus 4 leads 42.0 to 32.1.
- The biggest single-benchmark swing is SimpleBench: 58.8% for Claude Opus 4 and 27.2% for DeepSeek-V3.
- DeepSeek-V3 is cheaper at $0.24 / $0.90 per million input/output tokens, against $15 / $75 for Claude Opus 4.
- Claude Opus 4 accepts more context: 200K tokens versus 164K.
- DeepSeek-V3 has downloadable open weights; the other is API-only.

## Snapshot

| | Claude Opus 4 | DeepSeek-V3 |
|---|---|---|
| Provider | Anthropic | DeepSeek |
| Noometry Index | 43.1 | 39.5 |
| Rank | 100 | 166 |
| Context | 200K | 164K |
| Input $/M | $15 | $0.24 |
| Output $/M | $75 | $0.90 |
| Weights | Proprietary | Open |

## Coding

- Claude Opus 4: 47.2 (#62)
- DeepSeek-V3: 42.3 (#106)

| Benchmark | Claude Opus 4 | DeepSeek-V3 |
|---|---|---|
| Aider Polyglot | 72% | 55.1% |
| WeirdML | 43.7% | 36.1% |
| LMArena Coding | 1442 | 1368 |
| SWE-bench Verified | 70.7% | — |
| SWE-bench Verified (bash only) | 67.6% | — |
| SciCode | — | 35.8% |
| GSO | 6.9% | — |
| BigCodeBench Instruct | — | 50% |
| LiveBench Coding | — | 70.9% |
| BigCodeBench Complete | — | 62.2% |
| AlgoTune | 1.33 | — |
| HumanEval+ | — | 86.6% |
| MBPP+ | — | 73% |

## Agentic & Tool Use

- Claude Opus 4: 34.8 (#42)
- DeepSeek-V3: —

| Benchmark | Claude Opus 4 | DeepSeek-V3 |
|---|---|---|
| METR Time Horizons | 63.9% | 49.6% |
| Cybench | 38% | — |
| DeepResearch Bench | 46.8% | — |
| LMArena Search | 1127 | — |

## Reasoning

- Claude Opus 4: 27.3 (#121)
- DeepSeek-V3: 20.5 (#236)

| Benchmark | Claude Opus 4 | DeepSeek-V3 |
|---|---|---|
| SimpleBench | 58.8% | 27.2% |
| Kagi LLM Benchmark | 74.3% | 52.3% |
| CritPt | 0.3% | 0% |
| LMArena Hard Prompts | 1399 | 1365 |
| DTBench | 81.6% | 64.8% |
| LMCA | 37.4% | 15.5% |
| Epoch Capabilities Index | 142.67 | 135.94 |
| ForecastBench | 61.1 | 59.1 |
| ARC-AGI-2 | 8.6% | — |
| ARC-AGI-1 | 35.7% | — |
| EnigmaEval | 5.6% | — |
| LiveBench Reasoning | — | 65.8% |
| LiveBench Data Analysis | — | 60.9% |
| BIG-Bench Hard | — | 87.5% |
| HellaSwag | — | 88.9% |
| LiveBench | — | 66.9% |
| PIQA | — | 84.7% |
| WinoGrande | — | 85.2% |

## Math

- Claude Opus 4: 42.0 (#86)
- DeepSeek-V3: 32.1 (#219)

| Benchmark | Claude Opus 4 | DeepSeek-V3 |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 64.4% | 37.8% |
| Omni-MATH | 61.6% | 40.3% |
| LMArena Math | 1390 | 1373 |
| MATH Level 5 | 85% | 75.5% |
| FrontierMath (Feb 2025 set) | 4.5% | 1.7% |
| LiveBench Math | — | 73.5% |
| FrontierMath Tier 4 (v1) | 4.2% | — |

## Knowledge

- Claude Opus 4: 44.0 (#88)
- DeepSeek-V3: 37.5 (#155)

| Benchmark | Claude Opus 4 | DeepSeek-V3 |
|---|---|---|
| GPQA Diamond | 76.3% | 67.6% |
| MMLU-Pro | 87.5% | 72.3% |
| Confabulations | 15.9% | 26.1% |
| Vectara Hallucination Rate | 12% | 6.1% |
| GPQA (HELM) | 70.8% | 53.8% |
| LMArena Expert | 1386 | 1351 |
| Humanity's Last Exam | 10.7% | — |
| ARC (AI2) Challenge | — | 95.3% |
| MMLU | — | 87.2% |
| TriviaQA | — | 82.9% |

## Multimodal

- Claude Opus 4: 31.5 (#106)
- DeepSeek-V3: —

| Benchmark | Claude Opus 4 | DeepSeek-V3 |
|---|---|---|
| LMArena Vision | 1192 | — |
| GeoBench | 49% | — |
| VPCT | 38% | — |

## Multilingual

- Claude Opus 4: 48.8 (#138)
- DeepSeek-V3: 48.5 (#143)

| Benchmark | Claude Opus 4 | DeepSeek-V3 |
|---|---|---|
| LMArena Non-English | 1362 | 1358 |
| LMArena Chinese | 1386 | 1391 |
| LMArena French | 1372 | 1385 |
| LMArena German | 1391 | 1374 |
| LMArena Japanese | 1331 | 1333 |
| LMArena Korean | 1321 | 1319 |
| LMArena Russian | 1392 | 1373 |
| LMArena Spanish | 1389 | 1358 |

## Instruction Following

- Claude Opus 4: 77.1 (#28)
- DeepSeek-V3: 72.8 (#130)

| Benchmark | Claude Opus 4 | DeepSeek-V3 |
|---|---|---|
| IFEval | 91.8% | 83.2% |
| LMArena Instruction Following | 1406 | 1345 |
| LiveBench Instruction Following | — | 81.5% |

## Long Context

- Claude Opus 4: 39.6 (#172)
- DeepSeek-V3: 34.0 (#253)

| Benchmark | Claude Opus 4 | DeepSeek-V3 |
|---|---|---|
| Fiction.LiveBench | 61.1% | 50% |
| LMArena Longer Query | 1422 | 1352 |

## Writing & Preference

- Claude Opus 4: 61.2 (#89)
- DeepSeek-V3: 57.4 (#130)

| Benchmark | Claude Opus 4 | DeepSeek-V3 |
|---|---|---|
| LMArena Text | 1377 | 1375 |
| LMArena Creative Writing | 1387 | 1364 |
| Short-Story Creative Writing | 83.6% | 77% |
| EQ-Bench Creative Writing | 1580 | 1472 |
| WildBench | 85.2% | 83% |
| LMArena Multi-Turn | 1396 | 1389 |
| LiveBench Language | — | 49.1% |

## FAQ

### Is Claude Opus 4 better than DeepSeek-V3?

Claude Opus 4 is the stronger model overall, scoring 43.1 to 39.5 on the Noometry Index. DeepSeek-V3 costs 74× less per token, which makes it the better buy when Claude Opus 4's lead doesn't matter for your workload.

### Which is cheaper, Claude Opus 4 or DeepSeek-V3?

DeepSeek-V3 is cheaper. It lists at $0.24 per million input tokens and $0.90 per million output tokens; Claude Opus 4 lists at $15 and $75.

### Is Claude Opus 4 or DeepSeek-V3 better for coding?

Claude Opus 4 scores higher on coding benchmarks: 47.2 versus 42.3 in the Noometry coding category.

### Which has the bigger context window?

Claude Opus 4 does, with 200K tokens against 164K.

### How many benchmarks do Claude Opus 4 and DeepSeek-V3 share?

41 benchmarks have published results for both models. Claude Opus 4 has 56 scored results on Noometry and DeepSeek-V3 has 60.