# Claude Sonnet 4.5 vs DeepSeek V4 Flash

> DeepSeek V4 Flash is the stronger model overall, scoring 53.6 to 44.0 on the Noometry Index.

- Canonical page: https://noometry.com/compare/claude-sonnet-4-5-vs-deepseek-v4-flash
- Last updated: 2026-10-10
- Shared benchmarks: 39

## Summary

- They share 39 benchmarks with published results for both. Claude Sonnet 4.5 scores higher in 5 categories and DeepSeek V4 Flash in 3 categories; 5 gaps are clear of the uncertainty.
- The widest gap is in math, where DeepSeek V4 Flash leads 60.0 to 31.9.
- The biggest single-benchmark swing is NYT Connections (extended): 37.3% for Claude Sonnet 4.5 and 89.6% for DeepSeek V4 Flash.
- DeepSeek V4 Flash is cheaper at $0.15 / $0.60 per million input/output tokens, against $3 / $15 for Claude Sonnet 4.5.
- DeepSeek V4 Flash has downloadable open weights; the other is API-only.

## Snapshot

| | Claude Sonnet 4.5 | DeepSeek V4 Flash |
|---|---|---|
| Provider | Anthropic | DeepSeek |
| Noometry Index | 44.0 | 53.6 |
| Rank | 76 | 32 |
| Context | 1M | 1M |
| Input $/M | $3 | $0.15 |
| Output $/M | $15 | $0.60 |
| Weights | Proprietary | Open |

## Coding

- Claude Sonnet 4.5: 47.8 (#54)
- DeepSeek V4 Flash: 47.6 (#56)

| Benchmark | Claude Sonnet 4.5 | DeepSeek V4 Flash |
|---|---|---|
| LMArena WebDev | 1393 | 1581 |
| SciCode | 44.7% | 49.9% |
| WeirdML | 47.7% | 63% |
| LMArena Coding | 1489 | 1457 |
| ALE-Bench | 796.15 | 1,306 |
| SWE-bench Verified | 71.3% | — |
| FrontierCode | — | 18.8% |
| SWE-bench Verified (bash only) | 71.4% | — |
| SWE-bench Multilingual | 67% | — |
| GSO | 14.7% | — |
| AlgoTune | 1.52 | — |

## Agentic & Tool Use

- Claude Sonnet 4.5: 38.1 (#30)
- DeepSeek V4 Flash: —

| Benchmark | Claude Sonnet 4.5 | DeepSeek V4 Flash |
|---|---|---|
| Terminal-Bench | 46.5% | — |
| Berkeley Function Calling Leaderboard | 73.2% | — |
| GDPval | 42.5% | — |
| Remote Labor Index | 2.1% | — |
| τ²-bench Airline | 72% | — |
| τ²-bench Banking | 25.3% | — |
| τ²-bench Retail | 72.4% | — |
| τ²-bench Telecom | 84.9% | — |
| Cybench | 60% | — |
| DeepResearch Bench | 52.6% | — |
| OSWorld | 62.9% | — |
| LMArena Search | 1159 | — |
| METR Time Horizons | 67.4% | — |
| Vending-Bench 2 | 3,839 | — |

## Reasoning

- Claude Sonnet 4.5: 26.9 (#113)
- DeepSeek V4 Flash: 53.7 (#25)

| Benchmark | Claude Sonnet 4.5 | DeepSeek V4 Flash |
|---|---|---|
| ARC-AGI-2 | 13.6% | 61.4% |
| SimpleBench | 54.3% | 61.1% |
| Kagi LLM Benchmark | 57.9% | 52.2% |
| NYT Connections (extended) | 37.3% | 89.6% |
| ARC-AGI-1 | 63.7% | 89% |
| CritPt | 1.1% | 16.6% |
| Chess Puzzles | 12% | 33% |
| LMArena Hard Prompts | 1462 | 1443 |
| Mystery Game Puzzles | 17% | 34% |
| DTBench | 83.2% | 90.9% |
| LMCA | 38.8% | 41.7% |
| Epoch Capabilities Index | 146.84 | 154.51 |
| EnigmaEval | 6% | — |
| EBR-Bench | 2.4% | — |
| ForecastBench | 61.9 | — |

## Math

- Claude Sonnet 4.5: 31.9 (#190)
- DeepSeek V4 Flash: 60.0 (#31)

| Benchmark | Claude Sonnet 4.5 | DeepSeek V4 Flash |
|---|---|---|
| FrontierMath (Tiers 1-3) | 23.9% | 57.5% |
| FrontierMath Tier 4 | 2.4% | 24.4% |
| OTIS Mock AIME 2024-2025 | 77.8% | 94.4% |
| ProofBench | 19% | 56% |
| LMArena Math | 1448 | 1427 |
| MathArena Final-Answer Competitions | — | 76.5% |
| Omni-MATH | 55.3% | — |
| MATH Level 5 | 97.7% | — |
| FrontierMath (Feb 2025 set) | 15.2% | — |
| FrontierMath Tier 4 (v1) | 4.2% | — |

## Knowledge

- Claude Sonnet 4.5: 48.4 (#70)
- DeepSeek V4 Flash: 55.5 (#43)

| Benchmark | Claude Sonnet 4.5 | DeepSeek V4 Flash |
|---|---|---|
| GPQA Diamond | 82.3% | 91% |
| SimpleQA Verified | 30.7% | 33.6% |
| LMArena Expert | 1483 | 1441 |
| Humanity's Last Exam | 13.7% | — |
| MMLU-Pro | 86.9% | — |
| Vectara Hallucination Rate | 12% | — |
| GPQA (HELM) | 68.6% | — |

## Multimodal

- Claude Sonnet 4.5: 34.1 (#84)
- DeepSeek V4 Flash: —

| Benchmark | Claude Sonnet 4.5 | DeepSeek V4 Flash |
|---|---|---|
| VPCT | 39.8% | — |
| LMArena Document | 1450 | — |

## Multilingual

- Claude Sonnet 4.5: 53.6 (#65)
- DeepSeek V4 Flash: 53.2 (#71)

| Benchmark | Claude Sonnet 4.5 | DeepSeek V4 Flash |
|---|---|---|
| LMArena Non-English | 1425 | 1420 |
| LMArena Chinese | 1459 | 1467 |
| LMArena French | 1459 | 1438 |
| LMArena German | 1427 | 1418 |
| LMArena Japanese | 1390 | 1403 |
| LMArena Korean | 1403 | 1383 |
| LMArena Russian | 1436 | 1428 |
| LMArena Spanish | 1457 | 1437 |

## Instruction Following

- Claude Sonnet 4.5: 75.2 (#74)
- DeepSeek V4 Flash: 75.1 (#76)

| Benchmark | Claude Sonnet 4.5 | DeepSeek V4 Flash |
|---|---|---|
| LMArena Instruction Following | 1459 | 1421 |
| IFEval | 85% | — |

## Long Context

- Claude Sonnet 4.5: 45.2 (#45)
- DeepSeek V4 Flash: 43.8 (#81)

| Benchmark | Claude Sonnet 4.5 | DeepSeek V4 Flash |
|---|---|---|
| LMArena Longer Query | 1476 | 1435 |

## Writing & Preference

- Claude Sonnet 4.5: 66.7 (#33)
- DeepSeek V4 Flash: 64.0 (#59)

| Benchmark | Claude Sonnet 4.5 | DeepSeek V4 Flash |
|---|---|---|
| LMArena Text | 1439 | 1432 |
| LMArena Creative Writing | 1442 | 1403 |
| EQ-Bench Creative Writing | 1678 | 1559 |
| LMArena Multi-Turn | 1465 | 1449 |
| WildBench | 85.4% | — |

## FAQ

### Is Claude Sonnet 4.5 better than DeepSeek V4 Flash?

DeepSeek V4 Flash is the stronger model overall, scoring 53.6 to 44.0 on the Noometry Index.

### Which is cheaper, Claude Sonnet 4.5 or DeepSeek V4 Flash?

DeepSeek V4 Flash is cheaper. It lists at $0.15 per million input tokens and $0.60 per million output tokens; Claude Sonnet 4.5 lists at $3 and $15.

### Is Claude Sonnet 4.5 or DeepSeek V4 Flash better for coding?

They score almost the same on coding (47.8 vs 47.6); test both on your own repository before choosing.

### Which has the bigger context window?

Both accept 1M tokens.

### How many benchmarks do Claude Sonnet 4.5 and DeepSeek V4 Flash share?

39 benchmarks have published results for both models. Claude Sonnet 4.5 has 73 scored results on Noometry and DeepSeek V4 Flash has 41.