Model comparison
Claude Opus 5 vs DeepSeek V4 Pro
Claude Opus 5 is the stronger model overall, scoring 67.8 to 54.3 on the Noometry Index. DeepSeek V4 Pro costs 10× less per token, which makes it the better buy when Claude Opus 5's lead doesn't matter for your workload.
Last verified . 41 shared benchmarks.
DeepSeekSummary
- They share 41 benchmarks with published results for both. Claude Opus 5 scores higher in 9 categories and DeepSeek V4 Pro in 0 categories; 9 gaps are clear of the uncertainty.
- The widest gap is in agentic & tool use, where Claude Opus 5 leads 55.6 to 32.8.
- The biggest single-benchmark swing is ProofBench: 99% for Claude Opus 5 and 50% for DeepSeek V4 Pro.
- DeepSeek V4 Pro is cheaper at $0.66 / $1.98 per million input/output tokens, against $5 / $25 for Claude Opus 5.
- DeepSeek V4 Pro has downloadable open weights; the other is API-only.
Side by side
| Claude Opus 5 | DeepSeek V4 Pro | |
|---|---|---|
| Provider | Anthropic | DeepSeek |
| Noometry Index | 67.8 | 54.3 |
| Released | 2026-07-24 | 2026-04-24 |
| Weights | Proprietary | Open |
| Context window | 1M | 1M |
| Max output | 128K | 393K |
| Input $ / M tokens | $5 | $0.66 |
| Output $ / M tokens | $25 | $1.98 |
| Results tracked | 57 | 48 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Claude Opus 5 leads
Claude Opus 5: 67.5 (#5), DeepSeek V4 Pro: 52.4 (#34)
| Benchmark | Claude Opus 5 | DeepSeek V4 Pro |
|---|---|---|
| FrontierCode | 53.4% | 28.6% |
| LMArena WebDev | 1691 | 1582 |
| SciCode | 56.4% | 51% |
| WeirdML | 91.8% | 66.2% |
| LMArena Coding | 1534 | 1470 |
| ALE-Bench | 2,165 | 1,403 |
| SWE-bench Verified | — | 77.6% |
| DeepSWE | 73.6% | — |
| CursorBench | 46.6% | — |
| FrontierSWE | 52% | — |
Agentic & Tool Use Claude Opus 5 leads
Claude Opus 5: 55.6 (#1), DeepSeek V4 Pro: 32.8 (#58)
| Benchmark | Claude Opus 5 | DeepSeek V4 Pro |
|---|---|---|
| APEX-Agents | 65.8% | 47.3% |
| Vending-Bench 2 | 11,182 | 3,285 |
| OSWorld 2.0 | 31.4% | — |
| τ²-bench Banking | 48.7% | — |
| PostTrainBench | 35% | — |
| BALROG | 63.4% | — |
| GBAEval | 79.6% | — |
| GDP.pdf | 24% | — |
Reasoning Claude Opus 5 leads
Claude Opus 5: 77.2 (#4), DeepSeek V4 Pro: 56.5 (#24)
| Benchmark | Claude Opus 5 | DeepSeek V4 Pro |
|---|---|---|
| ARC-AGI-2 | 90.4% | 61.3% |
| NYT Connections (extended) | 94.3% | 91.3% |
| ARC-AGI-1 | 97.5% | 90.5% |
| CritPt | 29.1% | 18% |
| Chess Puzzles | 42% | 47% |
| LMArena Hard Prompts | 1526 | 1461 |
| Mystery Game Puzzles | 59% | 43% |
| DTBench | 97.9% | 93.9% |
| LMCA | 64.5% | 45.5% |
| Epoch Capabilities Index | 162.78 | 155.31 |
| SimpleBench | 80.6% | — |
| Kagi LLM Benchmark | — | 53.5% |
| EBR-Bench | 45.7% | — |
| Surface Evolver Bench | — | 40% |
| Bench to the Future 3 | 0.12 | — |
| ForecastBench | — | 56.1 |
Math Claude Opus 5 leads
Claude Opus 5: 86.2 (#8), DeepSeek V4 Pro: 64.8 (#30)
| Benchmark | Claude Opus 5 | DeepSeek V4 Pro |
|---|---|---|
| FrontierMath (Tiers 1-3) | 85.6% | 64.6% |
| FrontierMath Tier 4 | 73.2% | 26.8% |
| OTIS Mock AIME 2024-2025 | 98.9% | 98.6% |
| ProofBench | 99% | 50% |
| LMArena Math | 1531 | 1455 |
| MathArena Final-Answer Competitions | — | 76.6% |
Knowledge Claude Opus 5 leads
Claude Opus 5: 66.8 (#9), DeepSeek V4 Pro: 59.5 (#31)
| Benchmark | Claude Opus 5 | DeepSeek V4 Pro |
|---|---|---|
| GPQA Diamond | 93.9% | 91.7% |
| SimpleQA Verified | 59.9% | 52.9% |
| LMArena Expert | 1557 | 1464 |
| Vectara Hallucination Rate | — | 8.6% |
Multimodal Not comparable
Claude Opus 5: 50.8 (#8), DeepSeek V4 Pro: —
| Benchmark | Claude Opus 5 | DeepSeek V4 Pro |
|---|---|---|
| LMArena Vision | 1319 | — |
| Blueprint-Bench 2 | 30.4% | — |
| Furniture Assembly | 60.8% | — |
| LMArena Document | 1516 | — |
Multilingual Claude Opus 5 leads
Claude Opus 5: 58.8 (#4), DeepSeek V4 Pro: 54.4 (#45)
| Benchmark | Claude Opus 5 | DeepSeek V4 Pro |
|---|---|---|
| LMArena Non-English | 1501 | 1439 |
| LMArena Chinese | 1574 | 1486 |
| LMArena French | 1519 | 1472 |
| LMArena German | 1524 | 1458 |
| LMArena Japanese | 1516 | 1445 |
| LMArena Korean | 1521 | 1447 |
| LMArena Russian | 1507 | 1453 |
| LMArena Spanish | 1519 | 1458 |
Instruction Following Claude Opus 5 leads
Claude Opus 5: 79.2 (#7), DeepSeek V4 Pro: 76.1 (#47)
| Benchmark | Claude Opus 5 | DeepSeek V4 Pro |
|---|---|---|
| LMArena Instruction Following | 1517 | 1448 |
Long Context Claude Opus 5 leads
Claude Opus 5: 46.5 (#21), DeepSeek V4 Pro: 45.0 (#51)
| Benchmark | Claude Opus 5 | DeepSeek V4 Pro |
|---|---|---|
| LMArena Longer Query | 1515 | 1458 |
| CL-bench Life | — | 13.5% |
Writing & Preference Claude Opus 5 leads
Claude Opus 5: 79.2 (#1), DeepSeek V4 Pro: 65.5 (#46)
| Benchmark | Claude Opus 5 | DeepSeek V4 Pro |
|---|---|---|
| LMArena Text | 1507 | 1451 |
| LMArena Creative Writing | 1491 | 1446 |
| EQ-Bench Creative Writing | 2133 | 1553 |
| EQ-Bench 4 | 1385 | 1166 |
| LMArena Multi-Turn | 1499 | 1467 |
Frequently asked questions
Is Claude Opus 5 better than DeepSeek V4 Pro?
Claude Opus 5 is the stronger model overall, scoring 67.8 to 54.3 on the Noometry Index. DeepSeek V4 Pro costs 10× less per token, which makes it the better buy when Claude Opus 5's lead doesn't matter for your workload.
Which is cheaper, Claude Opus 5 or DeepSeek V4 Pro?
DeepSeek V4 Pro is cheaper. It lists at $0.66 per million input tokens and $1.98 per million output tokens; Claude Opus 5 lists at $5 and $25.
Is Claude Opus 5 or DeepSeek V4 Pro better for coding?
Claude Opus 5 scores higher on coding benchmarks: 67.5 versus 52.4 in the Noometry coding category.
Which has the bigger context window?
Both accept 1M tokens.
How many benchmarks do Claude Opus 5 and DeepSeek V4 Pro share?
41 benchmarks have published results for both models. Claude Opus 5 has 57 scored results on Noometry and DeepSeek V4 Pro has 48.