Model comparison
Claude 3 Opus vs DeepSeek V4 Pro
DeepSeek V4 Pro is the stronger model overall, scoring 54.3 to 29.5 on the Noometry Index.
Last verified . 26 shared benchmarks.
DeepSeekSummary
- They share 26 benchmarks with published results for both. Claude 3 Opus scores higher in 0 categories and DeepSeek V4 Pro in 9 categories; 9 gaps are clear of the uncertainty.
- The widest gap is in math, where DeepSeek V4 Pro leads 64.8 to 14.8.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 4.7% for Claude 3 Opus and 98.6% for DeepSeek V4 Pro.
- DeepSeek V4 Pro has downloadable open weights; the other is API-only.
Side by side
| Claude 3 Opus | DeepSeek V4 Pro | |
|---|---|---|
| Provider | Anthropic | DeepSeek |
| Noometry Index | 29.5 | 54.3 |
| Released | 2024-02-29 | 2026-04-24 |
| Weights | Proprietary | Open |
| Context window | — | 1M |
| Max output | — | 393K |
| Input $ / M tokens | — | $0.66 |
| Output $ / M tokens | — | $1.98 |
| Results tracked | 46 | 48 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding DeepSeek V4 Pro leads
Claude 3 Opus: 32.9 (#267), DeepSeek V4 Pro: 52.4 (#34)
| Benchmark | Claude 3 Opus | DeepSeek V4 Pro |
|---|---|---|
| WeirdML | 19.2% | 66.2% |
| LMArena Coding | 1264 | 1470 |
| SWE-bench Verified | — | 77.6% |
| FrontierCode | — | 28.6% |
| LMArena WebDev | — | 1582 |
| SciCode | — | 51% |
| BigCodeBench Instruct | 45.5% | — |
| LiveBench Coding | 38.6% | — |
| BigCodeBench Complete | 57.4% | — |
| ALE-Bench | — | 1,403 |
| HumanEval+ | 77.4% | — |
| MBPP+ | 73.3% | — |
Agentic & Tool Use DeepSeek V4 Pro leads
Claude 3 Opus: 24.6 (#116), DeepSeek V4 Pro: 32.8 (#58)
| Benchmark | Claude 3 Opus | DeepSeek V4 Pro |
|---|---|---|
| APEX-Agents | — | 47.3% |
| Cybench | 10% | — |
| METR Time Horizons | 29.5% | — |
| Vending-Bench 2 | — | 3,285 |
Reasoning DeepSeek V4 Pro leads
Claude 3 Opus: 14.6 (#324), DeepSeek V4 Pro: 56.5 (#24)
| Benchmark | Claude 3 Opus | DeepSeek V4 Pro |
|---|---|---|
| Chess Puzzles | 5% | 47% |
| LMArena Hard Prompts | 1245 | 1461 |
| DTBench | 61.6% | 93.9% |
| LMCA | 17% | 45.5% |
| Epoch Capabilities Index | 126.91 | 155.31 |
| ForecastBench | 58.4 | 56.1 |
| ARC-AGI-2 | — | 61.3% |
| SimpleBench | 23.5% | — |
| Kagi LLM Benchmark | — | 53.5% |
| NYT Connections (extended) | — | 91.3% |
| ARC-AGI-1 | — | 90.5% |
| CritPt | — | 18% |
| EnigmaEval | 0.8% | — |
| LiveBench Reasoning | 40.6% | — |
| Mystery Game Puzzles | — | 43% |
| LiveBench Data Analysis | 57.9% | — |
| Surface Evolver Bench | — | 40% |
| LiveBench | 49.2% | — |
| WinoGrande | 88.5% | — |
Math DeepSeek V4 Pro leads
Claude 3 Opus: 14.8 (#299), DeepSeek V4 Pro: 64.8 (#30)
| Benchmark | Claude 3 Opus | DeepSeek V4 Pro |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 4.7% | 98.6% |
| LMArena Math | 1273 | 1455 |
| FrontierMath (Tiers 1-3) | — | 64.6% |
| FrontierMath Tier 4 | — | 26.8% |
| MathArena Final-Answer Competitions | — | 76.6% |
| ProofBench | — | 50% |
| LiveBench Math | 43.6% | — |
| MATH Level 5 | 37.5% | — |
Knowledge DeepSeek V4 Pro leads
Claude 3 Opus: 24.5 (#267), DeepSeek V4 Pro: 59.5 (#31)
| Benchmark | Claude 3 Opus | DeepSeek V4 Pro |
|---|---|---|
| GPQA Diamond | 47.2% | 91.7% |
| SimpleQA Verified | 12.6% | 52.9% |
| LMArena Expert | 1223 | 1464 |
| Confabulations | 22.7% | — |
| Vectara Hallucination Rate | — | 8.6% |
| MMLU | 84.6% | — |
Multimodal Not comparable
Claude 3 Opus: 27.1 (#116), DeepSeek V4 Pro: —
| Benchmark | Claude 3 Opus | DeepSeek V4 Pro |
|---|---|---|
| LMArena Vision | 1023 | — |
Multilingual DeepSeek V4 Pro leads
Claude 3 Opus: 41.4 (#207), DeepSeek V4 Pro: 54.4 (#45)
| Benchmark | Claude 3 Opus | DeepSeek V4 Pro |
|---|---|---|
| LMArena Non-English | 1258 | 1439 |
| LMArena Chinese | 1248 | 1486 |
| LMArena French | 1275 | 1472 |
| LMArena German | 1258 | 1458 |
| LMArena Japanese | 1204 | 1445 |
| LMArena Korean | 1187 | 1447 |
| LMArena Russian | 1280 | 1453 |
| LMArena Spanish | 1246 | 1458 |
Instruction Following DeepSeek V4 Pro leads
Claude 3 Opus: 64.1 (#228), DeepSeek V4 Pro: 76.1 (#47)
| Benchmark | Claude 3 Opus | DeepSeek V4 Pro |
|---|---|---|
| LMArena Instruction Following | 1248 | 1448 |
| LiveBench Instruction Following | 63.9% | — |
Long Context DeepSeek V4 Pro leads
Claude 3 Opus: 38.2 (#202), DeepSeek V4 Pro: 45.0 (#51)
| Benchmark | Claude 3 Opus | DeepSeek V4 Pro |
|---|---|---|
| LMArena Longer Query | 1259 | 1458 |
| CL-bench Life | — | 13.5% |
Writing & Preference DeepSeek V4 Pro leads
Claude 3 Opus: 47.2 (#213), DeepSeek V4 Pro: 65.5 (#46)
| Benchmark | Claude 3 Opus | DeepSeek V4 Pro |
|---|---|---|
| LMArena Text | 1262 | 1451 |
| LMArena Creative Writing | 1235 | 1446 |
| LMArena Multi-Turn | 1275 | 1467 |
| EQ-Bench Creative Writing | — | 1553 |
| EQ-Bench 4 | — | 1166 |
| LiveBench Language | 50.4% | — |
Frequently asked questions
Is Claude 3 Opus better than DeepSeek V4 Pro?
DeepSeek V4 Pro is the stronger model overall, scoring 54.3 to 29.5 on the Noometry Index.
Is Claude 3 Opus or DeepSeek V4 Pro better for coding?
DeepSeek V4 Pro scores higher on coding benchmarks: 52.4 versus 32.9 in the Noometry coding category.
How many benchmarks do Claude 3 Opus and DeepSeek V4 Pro share?
26 benchmarks have published results for both models. Claude 3 Opus has 46 scored results on Noometry and DeepSeek V4 Pro has 48.