Model comparison
Claude 3 Opus vs DeepSeek-V3
DeepSeek-V3 is the stronger model overall, scoring 39.2 to 29.3 on the Noometry Index.
Last verified . 41 shared benchmarks.
DeepSeekSummary
- They share 41 benchmarks with published results for both. Claude 3 Opus scores higher in 1 category and DeepSeek-V3 in 7 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in math, where DeepSeek-V3 leads 30.4 to 13.9.
- The biggest single-benchmark swing is MATH Level 5: 37.5% for Claude 3 Opus and 75.5% for DeepSeek-V3.
- DeepSeek-V3 has downloadable open weights; the other is API-only.
Side by side
| Claude 3 Opus | DeepSeek-V3 | |
|---|---|---|
| Provider | Anthropic | DeepSeek |
| Noometry Index | 29.3 | 39.2 |
| Released | 2024-02-29 | 2024-12-26 |
| Weights | Proprietary | Open |
| Context window | — | 164K |
| Max output | — | 164K |
| Input $ / M tokens | — | $0.24 |
| Output $ / M tokens | — | $0.90 |
| Results tracked | 46 | 60 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding DeepSeek-V3 leads
Claude 3 Opus: 32.4 (#261), DeepSeek-V3: 41.9 (#99)
| Benchmark | Claude 3 Opus | DeepSeek-V3 |
|---|---|---|
| WeirdML | 19.2% | 36.1% |
| BigCodeBench Instruct | 45.5% | 50% |
| LiveBench Coding | 38.6% | 70.9% |
| LMArena Coding | 1264 | 1368 |
| BigCodeBench Complete | 57.4% | 62.2% |
| HumanEval+ | 77.4% | 86.6% |
| MBPP+ | 73.3% | 73% |
| Aider Polyglot | — | 55.1% |
| SciCode | — | 35.8% |
Agentic & Tool Use Not comparable
Claude 3 Opus: 24.3 (#110), DeepSeek-V3: —
| Benchmark | Claude 3 Opus | DeepSeek-V3 |
|---|---|---|
| METR Time Horizons | 29.5% | 49.6% |
| Cybench | 10% | — |
Reasoning DeepSeek-V3 leads
Claude 3 Opus: 14.2 (#318), DeepSeek-V3: 20.0 (#226)
| Benchmark | Claude 3 Opus | DeepSeek-V3 |
|---|---|---|
| SimpleBench | 23.5% | 27.2% |
| LiveBench Reasoning | 40.6% | 65.8% |
| LMArena Hard Prompts | 1246 | 1365 |
| DTBench | 61.6% | 64.8% |
| LiveBench Data Analysis | 57.9% | 60.9% |
| LMCA | 17% | 15.5% |
| Epoch Capabilities Index | 126.91 | 135.95 |
| ForecastBench | 58.4 | 59.1 |
| LiveBench | 49.2% | 66.9% |
| WinoGrande | 88.5% | 85.2% |
| Kagi LLM Benchmark | — | 52.3% |
| CritPt | — | 0% |
| Chess Puzzles | 5% | — |
| EnigmaEval | 0.8% | — |
| BIG-Bench Hard | — | 87.5% |
| HellaSwag | — | 88.9% |
| PIQA | — | 84.7% |
Math DeepSeek-V3 leads
Claude 3 Opus: 13.9 (#293), DeepSeek-V3: 30.4 (#208)
| Benchmark | Claude 3 Opus | DeepSeek-V3 |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 4.7% | 37.8% |
| LiveBench Math | 43.6% | 73.5% |
| LMArena Math | 1273 | 1373 |
| MATH Level 5 | 37.5% | 75.5% |
| Omni-MATH | — | 40.3% |
| FrontierMath (Feb 2025 set) | — | 1.7% |
Knowledge DeepSeek-V3 leads
Claude 3 Opus: 24.5 (#262), DeepSeek-V3: 37.5 (#147)
| Benchmark | Claude 3 Opus | DeepSeek-V3 |
|---|---|---|
| GPQA Diamond | 47.2% | 67.6% |
| Confabulations | 22.7% | 26.1% |
| LMArena Expert | 1223 | 1352 |
| MMLU | 84.6% | 87.2% |
| SimpleQA Verified | 12.6% | — |
| 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 Opus: 26.6 (#112), DeepSeek-V3: —
| Benchmark | Claude 3 Opus | DeepSeek-V3 |
|---|---|---|
| LMArena Vision | 1023 | — |
Multilingual DeepSeek-V3 leads
Claude 3 Opus: 41.6 (#202), DeepSeek-V3: 48.8 (#138)
| Benchmark | Claude 3 Opus | DeepSeek-V3 |
|---|---|---|
| LMArena Non-English | 1258 | 1358 |
| LMArena Chinese | 1248 | 1390 |
| LMArena French | 1274 | 1386 |
| LMArena German | 1258 | 1374 |
| LMArena Japanese | 1204 | 1333 |
| LMArena Korean | 1187 | 1321 |
| LMArena Russian | 1280 | 1372 |
| LMArena Spanish | 1246 | 1358 |
Instruction Following DeepSeek-V3 leads
Claude 3 Opus: 64.3 (#223), DeepSeek-V3: 73.0 (#125)
| Benchmark | Claude 3 Opus | DeepSeek-V3 |
|---|---|---|
| LiveBench Instruction Following | 63.9% | 81.5% |
| LMArena Instruction Following | 1248 | 1345 |
| IFEval | — | 83.2% |
Long Context Claude 3 Opus leads
Claude 3 Opus: 38.2 (#197), DeepSeek-V3: 34.0 (#248)
| Benchmark | Claude 3 Opus | DeepSeek-V3 |
|---|---|---|
| LMArena Longer Query | 1259 | 1353 |
| Fiction.LiveBench | — | 50% |
Writing & Preference DeepSeek-V3 leads
Claude 3 Opus: 47.4 (#208), DeepSeek-V3: 57.6 (#124)
| Benchmark | Claude 3 Opus | DeepSeek-V3 |
|---|---|---|
| LMArena Text | 1262 | 1375 |
| LMArena Creative Writing | 1235 | 1364 |
| LMArena Multi-Turn | 1275 | 1389 |
| LiveBench Language | 50.4% | 49.1% |
| Short-Story Creative Writing | — | 77% |
| EQ-Bench Creative Writing | — | 1472 |
| WildBench | — | 83% |
Frequently asked questions
Is Claude 3 Opus better than DeepSeek-V3?
DeepSeek-V3 is the stronger model overall, scoring 39.2 to 29.3 on the Noometry Index.
Is Claude 3 Opus or DeepSeek-V3 better for coding?
DeepSeek-V3 scores higher on coding benchmarks: 41.9 versus 32.4 in the Noometry coding category.
How many benchmarks do Claude 3 Opus and DeepSeek-V3 share?
41 benchmarks have published results for both models. Claude 3 Opus has 46 scored results on Noometry and DeepSeek-V3 has 60.