Model comparison
Claude 3 Haiku vs DeepSeek-V3
DeepSeek-V3 is the stronger model overall, scoring 39.5 to 25.9 on the Noometry Index.
Last verified . 34 shared benchmarks.
DeepSeekSummary
- They share 34 benchmarks with published results for both. Claude 3 Haiku scores higher in 1 category and DeepSeek-V3 in 7 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in writing & preference, where DeepSeek-V3 leads 57.4 to 29.7.
- The biggest single-benchmark swing is MATH Level 5: 14.9% for Claude 3 Haiku and 75.5% for DeepSeek-V3.
- DeepSeek-V3 has downloadable open weights; the other is API-only.
Side by side
| Claude 3 Haiku | DeepSeek-V3 | |
|---|---|---|
| Provider | Anthropic | DeepSeek |
| Noometry Index | 25.9 | 39.5 |
| Released | 2024-03-07 | 2024-12-26 |
| Weights | Proprietary | Open |
| Context window | — | 164K |
| Max output | — | 164K |
| Input $ / M tokens | — | $0.24 |
| Output $ / M tokens | — | $0.90 |
| Results tracked | 37 | 60 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding DeepSeek-V3 leads
Claude 3 Haiku: 26.4 (#325), DeepSeek-V3: 42.3 (#106)
| Benchmark | Claude 3 Haiku | DeepSeek-V3 |
|---|---|---|
| WeirdML | 9.8% | 36.1% |
| BigCodeBench Instruct | 39.4% | 50% |
| LMArena Coding | 1199 | 1368 |
| BigCodeBench Complete | 50.1% | 62.2% |
| HumanEval+ | 68.9% | 86.6% |
| MBPP+ | 68.8% | 73% |
| Aider Polyglot | — | 55.1% |
| SciCode | — | 35.8% |
| LiveBench Coding | — | 70.9% |
| CadEval | 12% | — |
Agentic & Tool Use Not comparable
Claude 3 Haiku: —, DeepSeek-V3: —
| Benchmark | Claude 3 Haiku | DeepSeek-V3 |
|---|---|---|
| METR Time Horizons | — | 49.6% |
Reasoning DeepSeek-V3 leads
Claude 3 Haiku: 16.3 (#307), DeepSeek-V3: 20.5 (#236)
| Benchmark | Claude 3 Haiku | DeepSeek-V3 |
|---|---|---|
| Kagi LLM Benchmark | 34.2% | 52.3% |
| LMArena Hard Prompts | 1174 | 1365 |
| DTBench | 50.1% | 64.8% |
| LMCA | 8.8% | 15.5% |
| Epoch Capabilities Index | 118.35 | 135.94 |
| ForecastBench | 53.2 | 59.1 |
| WinoGrande | 74.2% | 85.2% |
| SimpleBench | — | 27.2% |
| CritPt | — | 0% |
| LiveBench Reasoning | — | 65.8% |
| LiveBench Data Analysis | — | 60.9% |
| BIG-Bench Hard | — | 87.5% |
| HellaSwag | — | 88.9% |
| LiveBench | — | 66.9% |
| PIQA | — | 84.7% |
Math DeepSeek-V3 leads
Claude 3 Haiku: 9.8 (#319), DeepSeek-V3: 32.1 (#219)
| Benchmark | Claude 3 Haiku | DeepSeek-V3 |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 1.8% | 37.8% |
| LMArena Math | 1188 | 1373 |
| MATH Level 5 | 14.9% | 75.5% |
| Omni-MATH | — | 40.3% |
| LiveBench Math | — | 73.5% |
| FrontierMath (Feb 2025 set) | — | 1.7% |
Knowledge DeepSeek-V3 leads
Claude 3 Haiku: 17.3 (#285), DeepSeek-V3: 37.5 (#155)
| Benchmark | Claude 3 Haiku | DeepSeek-V3 |
|---|---|---|
| GPQA Diamond | 36.3% | 67.6% |
| Confabulations | 34.2% | 26.1% |
| LMArena Expert | 1148 | 1351 |
| MMLU | 73.8% | 87.2% |
| MMLU-Pro | — | 72.3% |
| Vectara Hallucination Rate | — | 6.1% |
| GPQA (HELM) | — | 53.8% |
| ARC (AI2) Challenge | — | 95.3% |
| TriviaQA | — | 82.9% |
Multimodal Not comparable
Claude 3 Haiku: 23.6 (#128), DeepSeek-V3: —
| Benchmark | Claude 3 Haiku | DeepSeek-V3 |
|---|---|---|
| LMArena Vision | 950 | — |
| ScienceQA | 72% | — |
Multilingual DeepSeek-V3 leads
Claude 3 Haiku: 36.0 (#243), DeepSeek-V3: 48.5 (#143)
| Benchmark | Claude 3 Haiku | DeepSeek-V3 |
|---|---|---|
| LMArena Non-English | 1178 | 1358 |
| LMArena Chinese | 1155 | 1391 |
| LMArena French | 1195 | 1385 |
| LMArena German | 1174 | 1374 |
| LMArena Japanese | 1102 | 1333 |
| LMArena Korean | 1109 | 1319 |
| LMArena Russian | 1204 | 1373 |
| LMArena Spanish | 1166 | 1358 |
Instruction Following DeepSeek-V3 leads
Claude 3 Haiku: 61.3 (#247), DeepSeek-V3: 72.8 (#130)
| Benchmark | Claude 3 Haiku | DeepSeek-V3 |
|---|---|---|
| LMArena Instruction Following | 1173 | 1345 |
| LiveBench Instruction Following | — | 81.5% |
| IFEval | — | 83.2% |
Long Context Claude 3 Haiku leads
Claude 3 Haiku: 36.1 (#237), DeepSeek-V3: 34.0 (#253)
| Benchmark | Claude 3 Haiku | DeepSeek-V3 |
|---|---|---|
| LMArena Longer Query | 1190 | 1352 |
| Fiction.LiveBench | — | 50% |
Writing & Preference DeepSeek-V3 leads
Claude 3 Haiku: 29.7 (#291), DeepSeek-V3: 57.4 (#130)
| Benchmark | Claude 3 Haiku | DeepSeek-V3 |
|---|---|---|
| LMArena Text | 1195 | 1375 |
| LMArena Creative Writing | 1157 | 1364 |
| EQ-Bench Creative Writing | 717 | 1472 |
| LMArena Multi-Turn | 1190 | 1389 |
| Short-Story Creative Writing | — | 77% |
| WildBench | — | 83% |
| LiveBench Language | — | 49.1% |
Frequently asked questions
Is Claude 3 Haiku better than DeepSeek-V3?
DeepSeek-V3 is the stronger model overall, scoring 39.5 to 25.9 on the Noometry Index.
Is Claude 3 Haiku or DeepSeek-V3 better for coding?
DeepSeek-V3 scores higher on coding benchmarks: 42.3 versus 26.4 in the Noometry coding category.
How many benchmarks do Claude 3 Haiku and DeepSeek-V3 share?
34 benchmarks have published results for both models. Claude 3 Haiku has 37 scored results on Noometry and DeepSeek-V3 has 60.