Model comparison
Claude Sonnet 5.5 vs DeepSeek V4 Pro
Claude Sonnet 5.5 is the stronger model overall, scoring 61.9 to 54.3 on the Noometry Index. DeepSeek V4 Pro costs 4.0× less per token, which makes it the better buy when Claude Sonnet 5.5's lead doesn't matter for your workload.
Last verified . 27 shared benchmarks.
DeepSeekSummary
- They share 27 benchmarks with published results for both. Claude Sonnet 5.5 scores higher in 8 categories and DeepSeek V4 Pro in 1 category; 6 gaps are clear of the uncertainty.
- The widest gap is in math, where Claude Sonnet 5.5 leads 87.9 to 64.8.
- The biggest single-benchmark swing is FrontierMath Tier 4: 80.5% for Claude Sonnet 5.5 and 26.8% for DeepSeek V4 Pro.
- DeepSeek V4 Pro is cheaper at $0.66 / $1.98 per million input/output tokens, against $2 / $10 for Claude Sonnet 5.5.
- DeepSeek V4 Pro has downloadable open weights; the other is API-only.
Side by side
| Claude Sonnet 5.5 | DeepSeek V4 Pro | |
|---|---|---|
| Provider | Anthropic | DeepSeek |
| Noometry Index | 61.9 | 54.3 |
| Released | 2026-09-28 | 2026-04-24 |
| Weights | Proprietary | Open |
| Context window | 1M | 1M |
| Max output | 128K | 393K |
| Input $ / M tokens | $2 | $0.66 |
| Output $ / M tokens | $10 | $1.98 |
| Results tracked | 32 | 48 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Claude Sonnet 5.5 leads
Claude Sonnet 5.5: 67.3 (#6), DeepSeek V4 Pro: 52.4 (#34)
| Benchmark | Claude Sonnet 5.5 | DeepSeek V4 Pro |
|---|---|---|
| FrontierCode | 52.1% | 28.6% |
| LMArena WebDev | 1774 | 1582 |
| SciCode | 61% | 51% |
| LMArena Coding | 1513 | 1470 |
| ALE-Bench | 1,819 | 1,403 |
| SWE-bench Verified | — | 77.6% |
| CursorBench | 55.5% | — |
| FrontierSWE | 61.9% | — |
| WeirdML | — | 66.2% |
Agentic & Tool Use Claude Sonnet 5.5 leads
Claude Sonnet 5.5: 45.0 (#16), DeepSeek V4 Pro: 32.8 (#58)
| Benchmark | Claude Sonnet 5.5 | DeepSeek V4 Pro |
|---|---|---|
| APEX-Agents | 75.5% | 47.3% |
| Vending-Bench 2 | — | 3,285 |
Reasoning DeepSeek V4 Pro leads
Claude Sonnet 5.5: 54.0 (#28), DeepSeek V4 Pro: 56.5 (#24)
| Benchmark | Claude Sonnet 5.5 | DeepSeek V4 Pro |
|---|---|---|
| NYT Connections (extended) | 80.5% | 91.3% |
| CritPt | 31.4% | 18% |
| LMArena Hard Prompts | 1495 | 1461 |
| Mystery Game Puzzles | 65% | 43% |
| Epoch Capabilities Index | 165.03 | 155.31 |
| ARC-AGI-2 | — | 61.3% |
| Kagi LLM Benchmark | — | 53.5% |
| ARC-AGI-1 | — | 90.5% |
| Chess Puzzles | — | 47% |
| DTBench | — | 93.9% |
| LMCA | — | 45.5% |
| Surface Evolver Bench | — | 40% |
| ForecastBench | — | 56.1 |
Math Claude Sonnet 5.5 leads
Claude Sonnet 5.5: 87.9 (#6), DeepSeek V4 Pro: 64.8 (#30)
| Benchmark | Claude Sonnet 5.5 | DeepSeek V4 Pro |
|---|---|---|
| FrontierMath (Tiers 1-3) | 88.8% | 64.6% |
| FrontierMath Tier 4 | 80.5% | 26.8% |
| OTIS Mock AIME 2024-2025 | 100% | 98.6% |
| ProofBench | 100% | 50% |
| LMArena Math | 1510 | 1455 |
| MathArena Final-Answer Competitions | — | 76.6% |
| FrontierMath Erdős | 2.9% | — |
Knowledge Claude Sonnet 5.5 leads
Claude Sonnet 5.5: 66.0 (#12), DeepSeek V4 Pro: 59.5 (#31)
| Benchmark | Claude Sonnet 5.5 | DeepSeek V4 Pro |
|---|---|---|
| GPQA Diamond | 95.6% | 91.7% |
| SimpleQA Verified | 46.5% | 52.9% |
| LMArena Expert | 1540 | 1464 |
| Vectara Hallucination Rate | — | 8.6% |
Multimodal Not comparable
Claude Sonnet 5.5: 51.5 (#6), DeepSeek V4 Pro: —
| Benchmark | Claude Sonnet 5.5 | DeepSeek V4 Pro |
|---|---|---|
| LMArena Vision | 1289 | — |
| Furniture Assembly | 75% | — |
Multilingual Too close to call
Claude Sonnet 5.5: 55.3 (#30), DeepSeek V4 Pro: 54.4 (#45)
| Benchmark | Claude Sonnet 5.5 | DeepSeek V4 Pro |
|---|---|---|
| LMArena Non-English | 1452 | 1439 |
| LMArena Chinese | 1522 | 1486 |
| LMArena Russian | 1451 | 1453 |
| LMArena French | — | 1472 |
| LMArena German | — | 1458 |
| LMArena Japanese | — | 1445 |
| LMArena Korean | — | 1447 |
| LMArena Spanish | — | 1458 |
Instruction Following Claude Sonnet 5.5 leads
Claude Sonnet 5.5: 78.3 (#11), DeepSeek V4 Pro: 76.1 (#47)
| Benchmark | Claude Sonnet 5.5 | DeepSeek V4 Pro |
|---|---|---|
| LMArena Instruction Following | 1495 | 1448 |
Long Context Too close to call
Claude Sonnet 5.5: 45.9 (#28), DeepSeek V4 Pro: 45.0 (#51)
| Benchmark | Claude Sonnet 5.5 | DeepSeek V4 Pro |
|---|---|---|
| LMArena Longer Query | 1498 | 1458 |
| CL-bench Life | — | 13.5% |
Writing & Preference Too close to call
Claude Sonnet 5.5: 66.0 (#40), DeepSeek V4 Pro: 65.5 (#46)
| Benchmark | Claude Sonnet 5.5 | DeepSeek V4 Pro |
|---|---|---|
| LMArena Text | 1471 | 1451 |
| LMArena Creative Writing | 1465 | 1446 |
| LMArena Multi-Turn | 1474 | 1467 |
| EQ-Bench Creative Writing | — | 1553 |
| EQ-Bench 4 | — | 1166 |
Frequently asked questions
Is Claude Sonnet 5.5 better than DeepSeek V4 Pro?
Claude Sonnet 5.5 is the stronger model overall, scoring 61.9 to 54.3 on the Noometry Index. DeepSeek V4 Pro costs 4.0× less per token, which makes it the better buy when Claude Sonnet 5.5's lead doesn't matter for your workload.
Which is cheaper, Claude Sonnet 5.5 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 Sonnet 5.5 lists at $2 and $10.
Is Claude Sonnet 5.5 or DeepSeek V4 Pro better for coding?
Claude Sonnet 5.5 scores higher on coding benchmarks: 67.3 versus 52.4 in the Noometry coding category.
Which has the bigger context window?
Both accept 1M tokens.
How many benchmarks do Claude Sonnet 5.5 and DeepSeek V4 Pro share?
27 benchmarks have published results for both models. Claude Sonnet 5.5 has 32 scored results on Noometry and DeepSeek V4 Pro has 48.