Model comparison
Claude 3.5 Sonnet vs DeepSeek V4 Flash
DeepSeek V4 Flash is the stronger model overall, scoring 53.6 to 34.6 on the Noometry Index.
Last verified . 24 shared benchmarks.
DeepSeekSummary
- They share 24 benchmarks with published results for both. Claude 3.5 Sonnet scores higher in 0 categories and DeepSeek V4 Flash in 8 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in math, where DeepSeek V4 Flash leads 60.3 to 19.2.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 8.5% for Claude 3.5 Sonnet and 94.4% for DeepSeek V4 Flash.
- DeepSeek V4 Flash has downloadable open weights; the other is API-only.
Side by side
| Claude 3.5 Sonnet | DeepSeek V4 Flash | |
|---|---|---|
| Provider | Anthropic | DeepSeek |
| Noometry Index | 34.6 | 53.6 |
| Released | 2024-06-20 | 2026-04-24 |
| Weights | Proprietary | Open |
| Context window | — | 1M |
| Max output | — | 393K |
| Input $ / M tokens | — | $0.15 |
| Output $ / M tokens | — | $0.60 |
| Results tracked | 60 | 41 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding DeepSeek V4 Flash leads
Claude 3.5 Sonnet: 39.0 (#165), DeepSeek V4 Flash: 47.9 (#59)
| Benchmark | Claude 3.5 Sonnet | DeepSeek V4 Flash |
|---|---|---|
| WeirdML | 40% | 63% |
| LMArena Coding | 1342 | 1457 |
| FrontierCode | — | 18.8% |
| Aider Polyglot | 51.6% | — |
| LMArena WebDev | — | 1582 |
| SciCode | — | 49.9% |
| GSO | 4.6% | — |
| BigCodeBench Instruct | 46.8% | — |
| LiveBench Coding | 67.1% | — |
| BigCodeBench Complete | 58.6% | — |
| CadEval | 48% | — |
| ALE-Bench | — | 1,306 |
| HumanEval+ | 81.7% | — |
| MBPP+ | 74.3% | — |
Agentic & Tool Use Not comparable
Claude 3.5 Sonnet: 32.3 (#67), DeepSeek V4 Flash: —
| Benchmark | Claude 3.5 Sonnet | DeepSeek V4 Flash |
|---|---|---|
| TheAgentCompany | 24% | — |
| Cybench | 17.5% | — |
| BALROG | 32.6% | — |
| METR Time Horizons | 45.2% | — |
Reasoning DeepSeek V4 Flash leads
Claude 3.5 Sonnet: 23.1 (#183), DeepSeek V4 Flash: 53.7 (#30)
| Benchmark | Claude 3.5 Sonnet | DeepSeek V4 Flash |
|---|---|---|
| SimpleBench | 41.4% | 61.1% |
| LMArena Hard Prompts | 1305 | 1444 |
| DTBench | 67.8% | 90.9% |
| Epoch Capabilities Index | 133.55 | 154.49 |
| ARC-AGI-2 | — | 61.4% |
| Kagi LLM Benchmark | — | 52.2% |
| NYT Connections (extended) | — | 89.6% |
| ARC-AGI-1 | — | 89% |
| CritPt | — | 16.6% |
| Chess Puzzles | — | 33% |
| EnigmaEval | 0.9% | — |
| LiveBench Reasoning | 56.7% | — |
| Mystery Game Puzzles | — | 34% |
| LiveBench Data Analysis | 55% | — |
| LMCA | — | 41.7% |
| ForecastBench | 60.7 | — |
| LiveBench | 59% | — |
Math DeepSeek V4 Flash leads
Claude 3.5 Sonnet: 19.2 (#288), DeepSeek V4 Flash: 60.3 (#37)
| Benchmark | Claude 3.5 Sonnet | DeepSeek V4 Flash |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 8.5% | 94.4% |
| LMArena Math | 1307 | 1427 |
| FrontierMath (Tiers 1-3) | — | 57.5% |
| FrontierMath Tier 4 | — | 24.4% |
| MathArena Final-Answer Competitions | — | 76.5% |
| ProofBench | — | 56% |
| Omni-MATH | 27.6% | — |
| LiveBench Math | 52.3% | — |
| MATH Level 5 | 56.9% | — |
| FrontierMath (Feb 2025 set) | 2.1% | — |
| FrontierMath Tier 4 (v1) | 0% | — |
Knowledge DeepSeek V4 Flash leads
Claude 3.5 Sonnet: 28.6 (#245), DeepSeek V4 Flash: 55.4 (#48)
| Benchmark | Claude 3.5 Sonnet | DeepSeek V4 Flash |
|---|---|---|
| GPQA Diamond | 55.3% | 91% |
| LMArena Expert | 1265 | 1441 |
| Humanity's Last Exam | 4.1% | — |
| SimpleQA Verified | — | 33.6% |
| MMLU-Pro | 77.7% | — |
| Confabulations | 19.9% | — |
| GPQA (HELM) | 56.5% | — |
| MMLU | 87.3% | — |
Multimodal Not comparable
Claude 3.5 Sonnet: 26.5 (#120), DeepSeek V4 Flash: —
| Benchmark | Claude 3.5 Sonnet | DeepSeek V4 Flash |
|---|---|---|
| LMArena Vision | 1125 | — |
| Video-MME | 60% | — |
| GeoBench | 62% | — |
| VPCT | 33% | — |
Multilingual DeepSeek V4 Flash leads
Claude 3.5 Sonnet: 43.2 (#185), DeepSeek V4 Flash: 53.0 (#72)
| Benchmark | Claude 3.5 Sonnet | DeepSeek V4 Flash |
|---|---|---|
| LMArena Non-English | 1283 | 1420 |
| LMArena Chinese | 1272 | 1468 |
| LMArena French | 1305 | 1439 |
| LMArena German | 1297 | 1418 |
| LMArena Japanese | 1234 | 1406 |
| LMArena Korean | 1200 | 1384 |
| LMArena Russian | 1306 | 1428 |
| LMArena Spanish | 1290 | 1436 |
Instruction Following DeepSeek V4 Flash leads
Claude 3.5 Sonnet: 68.8 (#182), DeepSeek V4 Flash: 74.9 (#81)
| Benchmark | Claude 3.5 Sonnet | DeepSeek V4 Flash |
|---|---|---|
| LMArena Instruction Following | 1297 | 1421 |
| LiveBench Instruction Following | 69.3% | — |
| IFEval | 85.5% | — |
Long Context DeepSeek V4 Flash leads
Claude 3.5 Sonnet: 39.9 (#167), DeepSeek V4 Flash: 43.8 (#85)
| Benchmark | Claude 3.5 Sonnet | DeepSeek V4 Flash |
|---|---|---|
| LMArena Longer Query | 1311 | 1434 |
Writing & Preference DeepSeek V4 Flash leads
Claude 3.5 Sonnet: 52.9 (#164), DeepSeek V4 Flash: 63.8 (#61)
| Benchmark | Claude 3.5 Sonnet | DeepSeek V4 Flash |
|---|---|---|
| LMArena Text | 1298 | 1432 |
| LMArena Creative Writing | 1292 | 1403 |
| EQ-Bench Creative Writing | 1451 | 1559 |
| LMArena Multi-Turn | 1326 | 1449 |
| Short-Story Creative Writing | 80.3% | — |
| WildBench | 79.2% | — |
| LiveBench Language | 53.8% | — |
Frequently asked questions
Is Claude 3.5 Sonnet better than DeepSeek V4 Flash?
DeepSeek V4 Flash is the stronger model overall, scoring 53.6 to 34.6 on the Noometry Index.
Is Claude 3.5 Sonnet or DeepSeek V4 Flash better for coding?
DeepSeek V4 Flash scores higher on coding benchmarks: 47.9 versus 39.0 in the Noometry coding category.
How many benchmarks do Claude 3.5 Sonnet and DeepSeek V4 Flash share?
24 benchmarks have published results for both models. Claude 3.5 Sonnet has 60 scored results on Noometry and DeepSeek V4 Flash has 41.