Claude Sonnet 5 vs DeepSeek V4 Pro 0813
Claude Sonnet 5 leads DeepSeek V4 Pro 0813 on coding, 57.7 to 56.1 on the Vibe Coding Index, and DeepSeek V4 Pro 0813 costs 5.1x less per output token ($1.98 vs $10.00 per million). Updated October 22, 2026; 74 benchmarks have a result for both.
Claude Sonnet 5 or DeepSeek V4 Pro 0813: which to pick
Claude Sonnet 5 and DeepSeek V4 Pro 0813 are level for coding, 57.7 against 56.1 on the Vibe Coding Index, so price and context decide.
Pick DeepSeek V4 Pro 0813 for output tokens 5.1x cheaper ($1.98 vs $10.00 per million) and a lower cost per solved task ($0.10 against $1.87). Pick Claude Sonnet 5 for work measured by AA Coding Index or Code migration (Vals AI), where it leads.
- Shared results
- 74
- Claude Sonnet 5 only
- 47
- DeepSeek V4 Pro 0813 only
- 4
- Like-for-like categories
- 7 of 9
Only benchmarks that tested both models count toward a pick; a result only one model has is shown but not compared.
Which one for your work
Each card is decided by one number both models carry. Where only one has it, the card says so and makes no call.
Coding
Vibe Coding Index, six public coding benchmarks
No call
Level: 57.7 against 56.1, inside 2 points.
6 and 5 benchmarks behind the scores
Intelligence
AA Intelligence Index
Claude Sonnet 5
Claude Sonnet 5 leads, 38.2 against 36.0.
Same board, both measured
Agentic work
AA Agentic Index
Claude Sonnet 5
Claude Sonnet 5 leads, 43.6 against 41.3.
Same board, both measured
Reasoning
GPQA Diamond
DeepSeek V4 Pro 0813
DeepSeek V4 Pro 0813 leads, 92.4% against 88.9%.
Same board, both measured
Knowledge
MMLU Pro
Claude Sonnet 5
Claude Sonnet 5 leads, 87.5% against 87.0%.
Same board, both measured
Web design
Design Arena: websites
No call
Only Claude Sonnet 5 has a Design Arena: websites result, 1274.
Directional only
General
Vals Index
No call
Only DeepSeek V4 Pro 0813 has a Vals Index result, 53.0%.
Directional only
Long documents
Listed context window
DeepSeek V4 Pro 0813
DeepSeek V4 Pro 0813 takes 1.05M tokens against 1M.
Listed by OpenRouter
Everyday cost
1K input + 500 output tokens
DeepSeek V4 Pro 0813
Claude Sonnet 5 $0.0070, DeepSeek V4 Pro 0813 $0.0016 per turn.
List prices
Cost per solved task
One SWE-bench Verified task, solved
DeepSeek V4 Pro 0813
Claude Sonnet 5 $1.87, DeepSeek V4 Pro 0813 $0.10.
Published attempt cost and pass rate
At a glance
- Coding index
- Claude Sonnet 5 57.7 from 6 benchmarks, DeepSeek V4 Pro 0813 56.1 from 5 benchmarks.
- Widest category
- Math, 7.5 points apart on average over 6 shared benchmarks.
- Widest single gap
- ProofBench: Claude Sonnet 5 77.0% against DeepSeek V4 Pro 0813 50.0%, 27.0 points apart on a 0 to 100 scale.
- Output price
- Claude Sonnet 5 $10.00 against DeepSeek V4 Pro 0813 $1.98 per million output tokens, 5.1x apart.
- Context window
- Claude Sonnet 5 takes 1M tokens, DeepSeek V4 Pro 0813 1.05M.
Benchmark by benchmark
First the score: boards each model leads, per category, where both have a result. Then every board, a category to a fold; bars are each result on a 0 to 100 scale, so compare the two in a row, not one row with another.
| Category | Claude Sonnet 5 | Bars | DeepSeek V4 Pro 0813 | Ties |
|---|---|---|---|---|
| Coding | 9 | 11 | 1 tie | |
| Intelligence | 2 | 0 | ||
| Agentic work | 7 | 1 | ||
| Reasoning | 4 | 5 | ||
| Math | 5 | 1 | ||
| Knowledge | 6 | 4 | ||
| General | 8 | 10 | ||
| All shared boards | 41 | 32 | 1 tie |
CodingLike-for-likeSix of these boards feed the Vibe Coding Index911
| Benchmark | Claude Sonnet 5 | DeepSeek V4 Pro 0813 | Leader |
|---|---|---|---|
| LiveBench coding | 74.7% | Not measured | Only Claude Sonnet 5 measured |
| LMArena WebDev | 1540 | 1582 | DeepSeek V4 Pro 0813 |
| AA Coding Index | 71.5 | 68.8 | Claude Sonnet 5 |
| LiveCodeBench (Vals AI) | 82.4% | 87.5% | DeepSeek V4 Pro 0813 |
| SWE-bench Verified (Vals AI) | 79.6% | 96.4% | DeepSeek V4 Pro 0813 |
| Terminal-Bench 4 (Vals AI) | 10.6% | 14.1% | DeepSeek V4 Pro 0813 |
| Vibe Code Bench (Vals AI) | 81.3% | 82.3% | DeepSeek V4 Pro 0813 |
| Code migration (Vals AI) | 44.4% | 41.5% | Claude Sonnet 5 |
| ProgramBench (Vals AI) | 0.0% | 0.0% | Tie |
| IOI (Vals AI) | 45.0% | 51.6% | DeepSeek V4 Pro 0813 |
| Terminal-Bench 2.1 (Vals AI) | 74.5% | 54.7% | Claude Sonnet 5 |
| Vibe Code Bench 1-100 (Vals AI) | 13.8% | 17.5% | DeepSeek V4 Pro 0813 |
| SkillsBench (Vals AI) | 46.5% | 53.8% | DeepSeek V4 Pro 0813 |
| CyberBench (Vals AI) | 70.3% | 72.9% | DeepSeek V4 Pro 0813 |
| SciCode | 54.3% | 51.0% | Claude Sonnet 5 |
| FrontierCode | 42.7% | 28.5% | Claude Sonnet 5 |
| GSO-Bench | 37.3% | Not measured | Only Claude Sonnet 5 measured |
| WeirdML | 68.8% | 66.2% | Claude Sonnet 5 |
| ALE-Bench | 1463.1 | 1403.2 | Claude Sonnet 5 |
| LMArena Text: coding | 1483 | 1469 | Claude Sonnet 5 |
| LMArena Text: software and IT | 1474 | 1462 | Claude Sonnet 5 |
| LMArena WebDev: React | 1541 | 1588 | DeepSeek V4 Pro 0813 |
| LMArena WebDev: HTML | 1527 | 1566 | DeepSeek V4 Pro 0813 |
| LMArena WebDev: from an image | 1540 | Not measured | Only Claude Sonnet 5 measured |
| SciCode | 10.7% | Not measured | Only Claude Sonnet 5 measured |
| Neither model has a result on the other 14 boards here | |||
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index20
| Benchmark | Claude Sonnet 5 | DeepSeek V4 Pro 0813 | Leader |
|---|---|---|---|
| AA Intelligence Index | 38.2 | 36.0 | Claude Sonnet 5 |
| Epoch Capabilities Index | 156.9 | 155.6 | Claude Sonnet 5 |
Agentic workLike-for-likePublic boards, shown as published, kept out of the index71
| Benchmark | Claude Sonnet 5 | DeepSeek V4 Pro 0813 | Leader |
|---|---|---|---|
| AA Agentic Index | 43.6 | 41.3 | Claude Sonnet 5 |
| Design Arena: agents building web apps | 1245 | Not measured | Only Claude Sonnet 5 measured |
| Design Arena: agents building full-stack apps | 1249 | Not measured | Only Claude Sonnet 5 measured |
| Design Arena: agents building mobile apps | 1215 | Not measured | Only Claude Sonnet 5 measured |
| Design Arena: agents building Android apps | 1227 | Not measured | Only Claude Sonnet 5 measured |
| Design Arena: agents building Godot games | 1225 | Not measured | Only Claude Sonnet 5 measured |
| Design Arena: agents building games | 1227 | Not measured | Only Claude Sonnet 5 measured |
| Design Arena: HTML slides | 1168 | Not measured | Only Claude Sonnet 5 measured |
| Design Arena: PowerPoint slides | 1191 | Not measured | Only Claude Sonnet 5 measured |
| Design Arena: spreadsheets | 1035 | Not measured | Only Claude Sonnet 5 measured |
| LiveBench research | 82.8% | Not measured | Only Claude Sonnet 5 measured |
| LiveBench browser | 72.0% | Not measured | Only Claude Sonnet 5 measured |
| LiveBench long agentic | 39.0% | Not measured | Only Claude Sonnet 5 measured |
| SREBench (Vals AI) | 0.4% | 1.9% | DeepSeek V4 Pro 0813 |
| Finance Agent (Vals AI) | 53.9% | 50.4% | Claude Sonnet 5 |
| Tax Agent Bench (Vals AI) | 29.4% | 26.8% | Claude Sonnet 5 |
| Legal Research Bench (Vals AI) | 41.8% | 40.9% | Claude Sonnet 5 |
| Harvey's Legal Agent Benchmark (Vals AI) | Not measured | 10.8% | Only DeepSeek V4 Pro 0813 measured |
| Excel Modeling Benchmark (Vals AI) | 66.3% | 52.8% | Claude Sonnet 5 |
| Terminal-Bench-Science (Vals AI) | 5.7% | 4.3% | Claude Sonnet 5 |
| APEX-Agents | 54.5% | 47.3% | Claude Sonnet 5 |
| Vending-Bench 2 | $6,378 | Not measured | Only Claude Sonnet 5 measured |
| Neither model has a result on the other 19 boards here | |||
ReasoningLike-for-likePublic boards, shown as published, kept out of the index45
| Benchmark | Claude Sonnet 5 | DeepSeek V4 Pro 0813 | Leader |
|---|---|---|---|
| GPQA Diamond | 88.9% | 92.4% | DeepSeek V4 Pro 0813 |
| LiveBench reasoning | 88.7% | Not measured | Only Claude Sonnet 5 measured |
| MMMU Pro (Vals AI) | 83.0% | Not measured | Only Claude Sonnet 5 measured |
| GPQA Diamond (Artificial Analysis run) | 91.1% | 92.8% | DeepSeek V4 Pro 0813 |
| Humanity's Last Exam (Artificial Analysis run) | 41.3% | 41.0% | Claude Sonnet 5 |
| GPQA Diamond (Epoch AI run) | 90.5% | 91.7% | DeepSeek V4 Pro 0813 |
| ARC-AGI | Not measured | 90.5% | Only DeepSeek V4 Pro 0813 measured |
| ARC-AGI-2 | Not measured | 61.3% | Only DeepSeek V4 Pro 0813 measured |
| SimpleBench | 60.6% | Not measured | Only Claude Sonnet 5 measured |
| Chess puzzles | 45.0% | 46.0% | DeepSeek V4 Pro 0813 |
| Mystery game puzzles | 35.0% | 43.0% | DeepSeek V4 Pro 0813 |
| LMArena Text: hard prompts | 1461 | 1458 | Claude Sonnet 5 |
| LMArena Text: expert | 1489 | 1459 | Claude Sonnet 5 |
| LMArena Text: hard prompts in English | 1464 | 1463 | Claude Sonnet 5 |
| LMArena Vision: diagrams | 1293 | Not measured | Only Claude Sonnet 5 measured |
| LMArena Vision: homework | 1311 | Not measured | Only Claude Sonnet 5 measured |
| Neither model has a result on the other 6 boards here | |||
MathLike-for-likePublic boards, shown as published, kept out of the index51
| Benchmark | Claude Sonnet 5 | DeepSeek V4 Pro 0813 | Leader |
|---|---|---|---|
| OTIS Mock AIME | 94.7% | 98.6% | DeepSeek V4 Pro 0813 |
| FrontierMath, tiers 1 to 3 | 65.6% | 64.6% | Claude Sonnet 5 |
| FrontierMath, tier 4 | 29.3% | 26.8% | Claude Sonnet 5 |
| ProofBench | 77.0% | 50.0% | Claude Sonnet 5 |
| LMArena Text: math | 1467 | 1441 | Claude Sonnet 5 |
| LMArena Text: mathematical work | 1481 | 1443 | Claude Sonnet 5 |
| Neither model has a result on the other 7 boards here | |||
KnowledgeLike-for-likePublic boards, shown as published, kept out of the index64
| Benchmark | Claude Sonnet 5 | DeepSeek V4 Pro 0813 | Leader |
|---|---|---|---|
| MMLU Pro | 87.5% | 87.0% | Claude Sonnet 5 |
| LegalBench (Vals AI) | 83.9% | 82.4% | Claude Sonnet 5 |
| CorpFin (Vals AI) | 67.0% | 65.4% | Claude Sonnet 5 |
| TaxEval (Vals AI) | 75.6% | 73.1% | Claude Sonnet 5 |
| MortgageTax (Vals AI) | 70.0% | Not measured | Only Claude Sonnet 5 measured |
| MedCode (Vals AI) | 47.5% | 42.5% | Claude Sonnet 5 |
| SimpleQA Verified | 33.7% | 52.9% | DeepSeek V4 Pro 0813 |
| LMArena Text: science | 1460 | 1457 | Claude Sonnet 5 |
| LMArena Text: medicine and healthcare | 1442 | 1446 | DeepSeek V4 Pro 0813 |
| LMArena Text: legal and government | 1454 | 1459 | DeepSeek V4 Pro 0813 |
| LMArena Text: business and finance | 1436 | 1442 | DeepSeek V4 Pro 0813 |
| Neither model has a result on the other 7 boards here | |||
Web designDirectional onlyPublic boards, shown as published, kept out of the index9 boards, none shared
| Benchmark | Claude Sonnet 5 | DeepSeek V4 Pro 0813 | Leader |
|---|---|---|---|
| Design Arena: websites | 1274 | Not measured | Only Claude Sonnet 5 measured |
| Design Arena: UI components | 1277 | Not measured | Only Claude Sonnet 5 measured |
| Design Arena: game dev | 1277 | Not measured | Only Claude Sonnet 5 measured |
| Design Arena: data viz | 1251 | Not measured | Only Claude Sonnet 5 measured |
| Design Arena: 3D | 1261 | Not measured | Only Claude Sonnet 5 measured |
| Design Arena: all code categories | 1275 | Not measured | Only Claude Sonnet 5 measured |
| Design Arena: SVG | 1192 | Not measured | Only Claude Sonnet 5 measured |
| Design Arena: ASCII art | 1207 | Not measured | Only Claude Sonnet 5 measured |
| LiveBench design | 82.0% | Not measured | Only Claude Sonnet 5 measured |
GeneralLike-for-likePublic boards, shown as published, kept out of the index810
| Benchmark | Claude Sonnet 5 | DeepSeek V4 Pro 0813 | Leader |
|---|---|---|---|
| Vals Index | Not measured | 53.0% | Only DeepSeek V4 Pro 0813 measured |
| LiveBench data analysis | 71.7% | Not measured | Only Claude Sonnet 5 measured |
| Public Benefits Bench (Vals AI) | 66.0% | Not measured | Only Claude Sonnet 5 measured |
| SAGE (Vals AI) | 48.9% | Not measured | Only Claude Sonnet 5 measured |
| MedScribe (Vals AI) | 76.0% | 80.2% | DeepSeek V4 Pro 0813 |
| Vals Multimodal Index | 68.8% | Not measured | Only Claude Sonnet 5 measured |
| Long context reasoning (AA-LCR) | 82.0% | 80.3% | Claude Sonnet 5 |
| DTBench | 92.5% | 93.9% | DeepSeek V4 Pro 0813 |
| LMCA | 50.0% | 45.5% | Claude Sonnet 5 |
| Surface Evolver Bench | 60.0% | Not measured | Only Claude Sonnet 5 measured |
| LMArena Text | 1442 | 1446 | DeepSeek V4 Pro 0813 |
| LMArena Text: instruction following | 1452 | 1448 | Claude Sonnet 5 |
| LMArena Text: creative writing | 1416 | 1439 | DeepSeek V4 Pro 0813 |
| LMArena Vision | 1273 | Not measured | Only Claude Sonnet 5 measured |
| LMArena Text: English | 1450 | 1457 | DeepSeek V4 Pro 0813 |
| LMArena Text: non-English | 1431 | 1431 | Claude Sonnet 5 |
| LMArena Text: Chinese | 1478 | 1481 | DeepSeek V4 Pro 0813 |
| LMArena Text: German | 1442 | Not measured | Only Claude Sonnet 5 measured |
| LMArena Text: Japanese | 1422 | Not measured | Only Claude Sonnet 5 measured |
| LMArena Text: Korean | 1412 | 1449 | DeepSeek V4 Pro 0813 |
| LMArena Text: French | 1463 | 1446 | Claude Sonnet 5 |
| LMArena Text: Spanish | 1438 | 1435 | Claude Sonnet 5 |
| LMArena Text: Russian | 1450 | 1448 | Claude Sonnet 5 |
| LMArena Text: Polish | 1425 | Not measured | Only Claude Sonnet 5 measured |
| LMArena Text: multi-turn | 1453 | 1454 | DeepSeek V4 Pro 0813 |
| LMArena Text: longer queries | 1462 | 1458 | Claude Sonnet 5 |
| LMArena Text: writing and language | 1436 | 1443 | DeepSeek V4 Pro 0813 |
| LMArena Text: entertainment and media | 1411 | 1424 | DeepSeek V4 Pro 0813 |
| LMArena Vision: English | 1275 | Not measured | Only Claude Sonnet 5 measured |
| LMArena Vision: Chinese | 1331 | Not measured | Only Claude Sonnet 5 measured |
| LMArena Vision: reading text | 1288 | Not measured | Only Claude Sonnet 5 measured |
| LMArena Vision: humor | 1261 | Not measured | Only Claude Sonnet 5 measured |
| LMArena Vision: creative writing | 1246 | Not measured | Only Claude Sonnet 5 measured |
| LMArena Document | 1466 | Not measured | Only Claude Sonnet 5 measured |
| Neither model has a result on the other 12 boards here | |||
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
| Benchmark | Claude Sonnet 5 | DeepSeek V4 Pro 0813 | Leader |
|---|---|---|---|
| Neither model has a result on 5 boards here | |||
What each workload costs
Three fixed token mixes at each model's listed price, so per-token rates become a bill.
Chat turn
1K input + 500 output tokens
DeepSeek V4 Pro 0813 costs less
Repository review
50K input + 3K output tokens
DeepSeek V4 Pro 0813 costs less
Agent loop
200K input + 10K output tokens
DeepSeek V4 Pro 0813 costs less
Listed OpenRouter rates, without cache or batch discounts. A workload that does not fit a model's context window would need splitting.
Specification differences
The leading value in a row is marked; a row with a missing value names no leader. Our two rows are marked in red.
| Specification | Anthropic Claude Sonnet 5 | DeepSeek V4 Pro 0813 |
|---|---|---|
| Maker | Anthropic | DeepSeek |
| Weights | No public weights listed | Open weights |
| Released | June 30, 2026 | August 12, 2026 |
| Context window | 1M | 1.05M |
| Input $/Mtok | $2.00 | $0.66 |
| Output $/Mtok | $10.00 | $1.98 |
| Coding index | 57.7 | 56.1 |
| Independent measurements behind it | 6 | 5 |
| Cost per solved task | $1.87 | $0.10 |
| Source | OpenRouter listing | OpenRouter listing |
Questions
Is Claude Sonnet 5 better than DeepSeek V4 Pro 0813 for coding?
Claude Sonnet 5 and DeepSeek V4 Pro 0813 are level for coding, 57.7 against 56.1 on the Vibe Coding Index, so price and context decide. Pick DeepSeek V4 Pro 0813 for output tokens 5.1x cheaper ($1.98 vs $10.00 per million) and a lower cost per solved task ($0.10 against $1.87). Pick Claude Sonnet 5 for work measured by AA Coding Index or Code migration (Vals AI), where it leads.
Which is cheaper, Claude Sonnet 5 or DeepSeek V4 Pro 0813?
Claude Sonnet 5 costs $2.00 per million input tokens and $10.00 per million output tokens; DeepSeek V4 Pro 0813 costs $0.66 and $1.98. Per solved SWE-bench Verified task, Claude Sonnet 5 costs $1.87 and DeepSeek V4 Pro 0813 $0.10.
Which has the bigger context window, Claude Sonnet 5 or DeepSeek V4 Pro 0813?
DeepSeek V4 Pro 0813, with 1.05M tokens against 1M.
How many benchmarks is the Claude Sonnet 5 vs DeepSeek V4 Pro 0813 comparison based on?
74 benchmarks have a result for both models as of October 22, 2026. Claude Sonnet 5's coding index rests on 6 independent benchmarks and DeepSeek V4 Pro 0813's on 5. Where only one model has a result, the page says so instead of naming a winner.