# Claude Opus 4.6 vs DeepSeek V4 Pro

> Claude Opus 4.6 is the stronger model overall, scoring 58.2 to 54.3 on the Noometry Index. DeepSeek V4 Pro costs 10× less per token, which makes it the better buy when Claude Opus 4.6's lead doesn't matter for your workload.

- Canonical page: https://noometry.com/compare/claude-opus-4-6-vs-deepseek-v4-pro
- Last updated: 2026-10-10
- Shared benchmarks: 45

## Summary

- They share 45 benchmarks with published results for both. Claude Opus 4.6 scores higher in 8 categories and DeepSeek V4 Pro in 1 category; 9 gaps are clear of the uncertainty.
- The widest gap is in agentic & tool use, where Claude Opus 4.6 leads 51.1 to 32.8.
- The biggest single-benchmark swing is Kagi LLM Benchmark: 83.6% for Claude Opus 4.6 and 53.5% for DeepSeek V4 Pro.
- DeepSeek V4 Pro is cheaper at $0.66 / $1.98 per million input/output tokens, against $5 / $25 for Claude Opus 4.6.
- DeepSeek V4 Pro has downloadable open weights; the other is API-only.

## Snapshot

| | Claude Opus 4.6 | DeepSeek V4 Pro |
|---|---|---|
| Provider | Anthropic | DeepSeek |
| Noometry Index | 58.2 | 54.3 |
| Rank | 20 | 31 |
| Context | 1M | 1M |
| Input $/M | $5 | $0.66 |
| Output $/M | $25 | $1.98 |
| Weights | Proprietary | Open |

## Coding

- Claude Opus 4.6: 57.2 (#20)
- DeepSeek V4 Pro: 52.4 (#34)

| Benchmark | Claude Opus 4.6 | DeepSeek V4 Pro |
|---|---|---|
| SWE-bench Verified | 78.7% | 77.6% |
| FrontierCode | 26.6% | 28.6% |
| LMArena WebDev | 1547 | 1582 |
| WeirdML | 78% | 66.2% |
| LMArena Coding | 1536 | 1470 |
| ALE-Bench | 996.5 | 1,403 |
| SWE-bench Verified (bash only) | 75.6% | — |
| SWE-bench Multilingual | 72% | — |
| SciCode | — | 51% |
| GSO | 41.2% | — |
| AlgoTune | 1.47 | — |

## Agentic & Tool Use

- Claude Opus 4.6: 51.1 (#4)
- DeepSeek V4 Pro: 32.8 (#58)

| Benchmark | Claude Opus 4.6 | DeepSeek V4 Pro |
|---|---|---|
| APEX-Agents | 46.3% | 47.3% |
| Vending-Bench 2 | 8,018 | 3,285 |
| Terminal-Bench | 79.8% | — |
| Remote Labor Index | 4.2% | — |
| τ²-bench Banking | 27.3% | — |
| Cybench | 93% | — |
| DeepResearch Bench | 55.3% | — |
| GBAEval | 44.1% | — |
| LMArena Search | 1253 | — |
| METR Time Horizons | 78.9% | — |

## Reasoning

- Claude Opus 4.6: 57.8 (#23)
- DeepSeek V4 Pro: 56.5 (#24)

| Benchmark | Claude Opus 4.6 | DeepSeek V4 Pro |
|---|---|---|
| ARC-AGI-2 | 69.2% | 61.3% |
| Kagi LLM Benchmark | 83.6% | 53.5% |
| NYT Connections (extended) | 92.1% | 91.3% |
| ARC-AGI-1 | 94% | 90.5% |
| Chess Puzzles | 17% | 47% |
| LMArena Hard Prompts | 1527 | 1461 |
| Mystery Game Puzzles | 25% | 43% |
| DTBench | 91.2% | 93.9% |
| LMCA | 55.8% | 45.5% |
| Epoch Capabilities Index | 155.24 | 155.31 |
| ForecastBench | 60 | 56.1 |
| SimpleBench | 67.6% | — |
| CritPt | — | 18% |
| EnigmaEval | 7.6% | — |
| Thematic Generalization | 80.6% | — |
| EBR-Bench | 12.7% | — |
| Surface Evolver Bench | — | 40% |

## Math

- Claude Opus 4.6: 63.0 (#31)
- DeepSeek V4 Pro: 64.8 (#30)

| Benchmark | Claude Opus 4.6 | DeepSeek V4 Pro |
|---|---|---|
| FrontierMath (Tiers 1-3) | 66% | 64.6% |
| FrontierMath Tier 4 | 26.8% | 26.8% |
| MathArena Final-Answer Competitions | 78.5% | 76.6% |
| OTIS Mock AIME 2024-2025 | 94.4% | 98.6% |
| ProofBench | 50% | 50% |
| LMArena Math | 1519 | 1455 |
| FrontierMath (Feb 2025 set) | 40.7% | — |
| FrontierMath Tier 4 (v1) | 22.9% | — |

## Knowledge

- Claude Opus 4.6: 61.9 (#26)
- DeepSeek V4 Pro: 59.5 (#31)

| Benchmark | Claude Opus 4.6 | DeepSeek V4 Pro |
|---|---|---|
| GPQA Diamond | 90.5% | 91.7% |
| SimpleQA Verified | 47% | 52.9% |
| Vectara Hallucination Rate | 12.2% | 8.6% |
| LMArena Expert | 1546 | 1464 |
| Humanity's Last Exam | 34.4% | — |

## Multimodal

- Claude Opus 4.6: 37.3 (#74)
- DeepSeek V4 Pro: —

| Benchmark | Claude Opus 4.6 | DeepSeek V4 Pro |
|---|---|---|
| LMArena Vision | 1316 | — |
| Furniture Assembly | 28.3% | — |
| LMArena Document | 1507 | — |

## Multilingual

- Claude Opus 4.6: 57.9 (#6)
- DeepSeek V4 Pro: 54.4 (#45)

| Benchmark | Claude Opus 4.6 | DeepSeek V4 Pro |
|---|---|---|
| LMArena Non-English | 1489 | 1439 |
| LMArena Chinese | 1551 | 1486 |
| LMArena French | 1513 | 1472 |
| LMArena German | 1502 | 1458 |
| LMArena Japanese | 1484 | 1445 |
| LMArena Korean | 1464 | 1447 |
| LMArena Russian | 1497 | 1453 |
| LMArena Spanish | 1510 | 1458 |

## Instruction Following

- Claude Opus 4.6: 79.5 (#4)
- DeepSeek V4 Pro: 76.1 (#47)

| Benchmark | Claude Opus 4.6 | DeepSeek V4 Pro |
|---|---|---|
| LMArena Instruction Following | 1523 | 1448 |

## Long Context

- Claude Opus 4.6: 48.1 (#13)
- DeepSeek V4 Pro: 45.0 (#51)

| Benchmark | Claude Opus 4.6 | DeepSeek V4 Pro |
|---|---|---|
| CL-bench Life | 17% | 13.5% |
| LMArena Longer Query | 1520 | 1458 |
| CL-bench | 20.7% | — |

## Writing & Preference

- Claude Opus 4.6: 73.5 (#10)
- DeepSeek V4 Pro: 65.5 (#46)

| Benchmark | Claude Opus 4.6 | DeepSeek V4 Pro |
|---|---|---|
| LMArena Text | 1503 | 1451 |
| LMArena Creative Writing | 1505 | 1446 |
| EQ-Bench Creative Writing | 1809 | 1553 |
| EQ-Bench 4 | 1223 | 1166 |
| LMArena Multi-Turn | 1513 | 1467 |

## FAQ

### Is Claude Opus 4.6 better than DeepSeek V4 Pro?

Claude Opus 4.6 is the stronger model overall, scoring 58.2 to 54.3 on the Noometry Index. DeepSeek V4 Pro costs 10× less per token, which makes it the better buy when Claude Opus 4.6's lead doesn't matter for your workload.

### Which is cheaper, Claude Opus 4.6 or DeepSeek V4 Pro?

DeepSeek V4 Pro is cheaper. It lists at $0.66 per million input tokens and $1.98 per million output tokens; Claude Opus 4.6 lists at $5 and $25.

### Is Claude Opus 4.6 or DeepSeek V4 Pro better for coding?

Claude Opus 4.6 scores higher on coding benchmarks: 57.2 versus 52.4 in the Noometry coding category.

### Which has the bigger context window?

Both accept 1M tokens.

### How many benchmarks do Claude Opus 4.6 and DeepSeek V4 Pro share?

45 benchmarks have published results for both models. Claude Opus 4.6 has 68 scored results on Noometry and DeepSeek V4 Pro has 48.