# Claude 3.5 Haiku vs DeepSeek-V3

> DeepSeek-V3 is the stronger model overall, scoring 39.2 to 28.9 on the Noometry Index.

- Canonical page: https://noometry.com/compare/claude-3-5-haiku-vs-deepseek-v3
- Last updated: 2026-10-10
- Shared benchmarks: 45

## Summary

- They share 45 benchmarks with published results for both. Claude 3.5 Haiku scores higher in 1 category and DeepSeek-V3 in 7 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in knowledge, where DeepSeek-V3 leads 37.5 to 18.7.
- The biggest single-benchmark swing is LiveBench Math: 35.5% for Claude 3.5 Haiku and 73.5% for DeepSeek-V3.
- DeepSeek-V3 has downloadable open weights; the other is API-only.

## Snapshot

| | Claude 3.5 Haiku | DeepSeek-V3 |
|---|---|---|
| Provider | Anthropic | DeepSeek |
| Noometry Index | 28.9 | 39.2 |
| Rank | 309 | 157 |
| Context | — | 164K |
| Input $/M | — | $0.24 |
| Output $/M | — | $0.90 |
| Weights | Proprietary | Open |

## Coding

- Claude 3.5 Haiku: 32.6 (#258)
- DeepSeek-V3: 41.9 (#99)

| Benchmark | Claude 3.5 Haiku | DeepSeek-V3 |
|---|---|---|
| Aider Polyglot | 28% | 55.1% |
| SciCode | 27.4% | 35.8% |
| WeirdML | 30.7% | 36.1% |
| BigCodeBench Instruct | 46.1% | 50% |
| LiveBench Coding | 51.4% | 70.9% |
| LMArena Coding | 1287 | 1368 |
| BigCodeBench Complete | 59% | 62.2% |
| CadEval | 32% | — |
| HumanEval+ | — | 86.6% |
| MBPP+ | — | 73% |

## Agentic & Tool Use

- Claude 3.5 Haiku: 27.7 (#90)
- DeepSeek-V3: —

| Benchmark | Claude 3.5 Haiku | DeepSeek-V3 |
|---|---|---|
| BALROG | 19.3% | — |
| METR Time Horizons | — | 49.6% |

## Reasoning

- Claude 3.5 Haiku: 17.1 (#286)
- DeepSeek-V3: 20.0 (#226)

| Benchmark | Claude 3.5 Haiku | DeepSeek-V3 |
|---|---|---|
| CritPt | 0% | 0% |
| LiveBench Reasoning | 28.1% | 65.8% |
| LMArena Hard Prompts | 1251 | 1365 |
| DTBench | 56.7% | 64.8% |
| LiveBench Data Analysis | 48.5% | 60.9% |
| Epoch Capabilities Index | 127.15 | 135.95 |
| LiveBench | 43.5% | 66.9% |
| SimpleBench | — | 27.2% |
| Kagi LLM Benchmark | — | 52.3% |
| LMCA | — | 15.5% |
| BIG-Bench Hard | — | 87.5% |
| ForecastBench | — | 59.1 |
| HellaSwag | — | 88.9% |
| PIQA | — | 84.7% |
| WinoGrande | — | 85.2% |

## Math

- Claude 3.5 Haiku: 13.7 (#294)
- DeepSeek-V3: 30.4 (#208)

| Benchmark | Claude 3.5 Haiku | DeepSeek-V3 |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 4.3% | 37.8% |
| Omni-MATH | 22.4% | 40.3% |
| LiveBench Math | 35.5% | 73.5% |
| LMArena Math | 1244 | 1373 |
| MATH Level 5 | 46.4% | 75.5% |
| FrontierMath (Feb 2025 set) | 0.3% | 1.7% |

## Knowledge

- Claude 3.5 Haiku: 18.7 (#276)
- DeepSeek-V3: 37.5 (#147)

| Benchmark | Claude 3.5 Haiku | DeepSeek-V3 |
|---|---|---|
| GPQA Diamond | 38.1% | 67.6% |
| MMLU-Pro | 60.5% | 72.3% |
| Confabulations | 36.7% | 26.1% |
| GPQA (HELM) | 36.3% | 53.8% |
| LMArena Expert | 1208 | 1352 |
| MMLU | 74.3% | 87.2% |
| Vectara Hallucination Rate | — | 6.1% |
| ARC (AI2) Challenge | — | 95.3% |
| TriviaQA | — | 82.9% |

## Multimodal

- Claude 3.5 Haiku: 26.3 (#113)
- DeepSeek-V3: —

| Benchmark | Claude 3.5 Haiku | DeepSeek-V3 |
|---|---|---|
| LMArena Vision | 1092 | — |
| GeoBench | 34% | — |

## Multilingual

- Claude 3.5 Haiku: 40.2 (#213)
- DeepSeek-V3: 48.8 (#138)

| Benchmark | Claude 3.5 Haiku | DeepSeek-V3 |
|---|---|---|
| LMArena Non-English | 1238 | 1358 |
| LMArena Chinese | 1229 | 1390 |
| LMArena French | 1262 | 1386 |
| LMArena German | 1237 | 1374 |
| LMArena Japanese | 1175 | 1333 |
| LMArena Korean | 1172 | 1321 |
| LMArena Russian | 1253 | 1372 |
| LMArena Spanish | 1261 | 1358 |

## Instruction Following

- Claude 3.5 Haiku: 63.1 (#229)
- DeepSeek-V3: 73.0 (#125)

| Benchmark | Claude 3.5 Haiku | DeepSeek-V3 |
|---|---|---|
| LiveBench Instruction Following | 61.9% | 81.5% |
| IFEval | 79.2% | 83.2% |
| LMArena Instruction Following | 1241 | 1345 |

## Long Context

- Claude 3.5 Haiku: 38.3 (#195)
- DeepSeek-V3: 34.0 (#248)

| Benchmark | Claude 3.5 Haiku | DeepSeek-V3 |
|---|---|---|
| LMArena Longer Query | 1261 | 1353 |
| Fiction.LiveBench | — | 50% |

## Writing & Preference

- Claude 3.5 Haiku: 42.9 (#228)
- DeepSeek-V3: 57.6 (#124)

| Benchmark | Claude 3.5 Haiku | DeepSeek-V3 |
|---|---|---|
| LMArena Text | 1255 | 1375 |
| LMArena Creative Writing | 1233 | 1364 |
| Short-Story Creative Writing | 73.5% | 77% |
| EQ-Bench Creative Writing | 1146 | 1472 |
| WildBench | 76% | 83% |
| LMArena Multi-Turn | 1265 | 1389 |
| LiveBench Language | 35.4% | 49.1% |

## FAQ

### Is Claude 3.5 Haiku better than DeepSeek-V3?

DeepSeek-V3 is the stronger model overall, scoring 39.2 to 28.9 on the Noometry Index.

### Is Claude 3.5 Haiku or DeepSeek-V3 better for coding?

DeepSeek-V3 scores higher on coding benchmarks: 41.9 versus 32.6 in the Noometry coding category.

### How many benchmarks do Claude 3.5 Haiku and DeepSeek-V3 share?

45 benchmarks have published results for both models. Claude 3.5 Haiku has 49 scored results on Noometry and DeepSeek-V3 has 60.