Claude Opus 5 vs DeepSeek V4 Pro 0423
Updated October 21, 2026. 80 benchmarks have a result for both. Coding is ranked on the Vibe Coding Index; every other board is shown as published.
Decision reading
Pick Claude Opus 5 for coding: it scores 81.8 on the Vibe Coding Index against 35.3 for DeepSeek V4 Pro 0423.
Pick DeepSeek V4 Pro 0423 when output tokens cost 13.1x less, or when a solved task costs $0.57 against $1.33.
- Shared results
- 80
- Claude Opus 5 only
- 50
- DeepSeek V4 Pro 0423 only
- 3
- Like-for-like categories
- 8 of 9
Shared results can support a head-to-head reading. Results only one model has describe coverage, not superiority.
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, four public coding benchmarks
Claude Opus 5
Claude Opus 5 scores 81.8 against 35.3.
2 and 1 benchmarks behind the scores
Intelligence
AA Intelligence Index
Claude Opus 5
Claude Opus 5 leads, 50.8 against 30.4.
Same board, both measured
Agentic work
AA Agentic Index
Claude Opus 5
Claude Opus 5 leads, 56.5 against 26.3.
Same board, both measured
Reasoning
GPQA Diamond
Claude Opus 5
Claude Opus 5 leads, 93.4% against 89.4%.
Same board, both measured
Knowledge
MMLU Pro
Claude Opus 5
Claude Opus 5 leads, 91.6% against 87.3%.
Same board, both measured
Web design
Design Arena: websites
Claude Opus 5
Claude Opus 5 leads, 1309 against 1235.
Same board, both measured
Long documents
Listed context window
DeepSeek V4 Pro 0423
DeepSeek V4 Pro 0423 takes 1.05M tokens against 1M.
Listed by OpenRouter
Everyday cost
1K input + 500 output tokens
DeepSeek V4 Pro 0423
Claude Opus 5 $0.018, DeepSeek V4 Pro 0423 $0.0019 per turn.
List prices
Cost per solved task
One SWE-bench Verified task, solved
DeepSeek V4 Pro 0423
Claude Opus 5 $1.33, DeepSeek V4 Pro 0423 $0.57.
Published attempt cost and pass rate
At a glance
- Coding index
- Claude Opus 5 81.8 from 2 benchmarks, DeepSeek V4 Pro 0423 35.3 from 1 benchmark.
- Widest category
- Math, 37.7 points apart on average over 6 shared benchmarks.
- Widest single gap
- ProofBench: Claude Opus 5 99.0% against DeepSeek V4 Pro 0423 16.0%, 83.0 points apart on a 0 to 100 scale.
- Output price
- Claude Opus 5 $25.00 against DeepSeek V4 Pro 0423 $1.9105 per million output tokens, 13.1x apart.
- Context window
- Claude Opus 5 takes 1M tokens, DeepSeek V4 Pro 0423 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 Opus 5 | Bars | DeepSeek V4 Pro 0423 | Ties |
|---|---|---|---|---|
| Coding | 17 | 0 | ||
| Intelligence | 2 | 0 | ||
| Agentic work | 9 | 0 | ||
| Reasoning | 7 | 0 | ||
| Math | 6 | 0 | ||
| Knowledge | 10 | 0 | ||
| Web design | 8 | 0 | ||
| General | 21 | 0 | ||
| All shared boards | 80 | 0 |
CodingLike-for-likeFour of these boards feed the Vibe Coding Index170
| Benchmark | Claude Opus 5 | DeepSeek V4 Pro 0423 | Leader |
|---|---|---|---|
| LiveBench coding | 90.8% | Not measured | Only Claude Opus 5 measured |
| LMArena WebDev | 1689 | 1447 | Claude Opus 5 |
| AA Coding Index | 78.0 | 59.4 | Claude Opus 5 |
| LiveCodeBench (Vals AI) | 89.0% | 87.5% | Claude Opus 5 |
| SWE-bench Verified (Vals AI) | 97.0% | 77.4% | Claude Opus 5 |
| Terminal-Bench 4 (Vals AI) | 53.5% | 11.1% | Claude Opus 5 |
| Vibe Code Bench (Vals AI) | 88.4% | 49.9% | Claude Opus 5 |
| Code migration (Vals AI) | 57.5% | 26.2% | Claude Opus 5 |
| ProgramBench (Vals AI) | 3.0% | 0.0% | Claude Opus 5 |
| IOI (Vals AI) | 84.3% | Not measured | Only Claude Opus 5 measured |
| Terminal-Bench 2.1 (Vals AI) | 84.6% | 50.2% | Claude Opus 5 |
| Vibe Code Bench 1-100 (Vals AI) | 28.5% | Not measured | Only Claude Opus 5 measured |
| SkillsBench (Vals AI) | 60.4% | 51.3% | Claude Opus 5 |
| CyberBench (Vals AI) | 65.4% | Not measured | Only Claude Opus 5 measured |
| SciCode | 56.4% | Not measured | Only Claude Opus 5 measured |
| FrontierCode | 53.4% | 17.6% | Claude Opus 5 |
| FrontierSWE | 52.0% | Not measured | Only Claude Opus 5 measured |
| WeirdML | 91.8% | 48.9% | Claude Opus 5 |
| ALE-Bench | 2164.6 | 1006.1 | Claude Opus 5 |
| LMArena Text: coding | 1533 | 1470 | Claude Opus 5 |
| LMArena Text: software and IT | 1527 | 1469 | Claude Opus 5 |
| LMArena WebDev: React | 1711 | 1441 | Claude Opus 5 |
| LMArena WebDev: HTML | 1680 | 1435 | Claude Opus 5 |
| LMArena WebDev: from an image | 1662 | Not measured | Only Claude Opus 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 Opus 5 | DeepSeek V4 Pro 0423 | Leader |
|---|---|---|---|
| AA Intelligence Index | 50.8 | 30.4 | Claude Opus 5 |
| Epoch Capabilities Index | 163.9 | 149.1 | Claude Opus 5 |
Agentic workLike-for-likePublic boards, shown as published, kept out of the index90
| Benchmark | Claude Opus 5 | DeepSeek V4 Pro 0423 | Leader |
|---|---|---|---|
| AA Agentic Index | 56.5 | 26.3 | Claude Opus 5 |
| Design Arena: agents building web apps | 1266 | 1000 | Claude Opus 5 |
| Design Arena: agents building full-stack apps | 1306 | 948 | Claude Opus 5 |
| Design Arena: agents building mobile apps | 1348 | Not measured | Only Claude Opus 5 measured |
| Design Arena: agents building Android apps | 1258 | Not measured | Only Claude Opus 5 measured |
| Design Arena: agents building Godot games | 1279 | 1059 | Claude Opus 5 |
| Design Arena: agents building games | 1261 | Not measured | Only Claude Opus 5 measured |
| Design Arena: PowerPoint slides | 1265 | Not measured | Only Claude Opus 5 measured |
| LiveBench research | 88.9% | Not measured | Only Claude Opus 5 measured |
| LiveBench browser | 84.0% | Not measured | Only Claude Opus 5 measured |
| LiveBench long agentic | 62.1% | Not measured | Only Claude Opus 5 measured |
| SREBench (Vals AI) | 12.2% | Not measured | Only Claude Opus 5 measured |
| Finance Agent (Vals AI) | 58.6% | 44.1% | Claude Opus 5 |
| Tax Agent Bench (Vals AI) | 45.5% | 25.2% | Claude Opus 5 |
| Legal Research Bench (Vals AI) | 55.3% | 23.1% | Claude Opus 5 |
| Excel Modeling Benchmark (Vals AI) | 73.6% | 51.6% | Claude Opus 5 |
| BioMysteryBench (Vals AI) | 79.3% | Not measured | Only Claude Opus 5 measured |
| Terminal-Bench-Science (Vals AI) | 27.1% | Not measured | Only Claude Opus 5 measured |
| Vals Science Index | 47.9% | Not measured | Only Claude Opus 5 measured |
| Time Horizon Index (Vals AI) | 18.8% | Not measured | Only Claude Opus 5 measured |
| CUA-bench (Vals AI) | 9.0% | Not measured | Only Claude Opus 5 measured |
| APEX-Agents | 65.8% | Not measured | Only Claude Opus 5 measured |
| OSWorld 2.0 | 31.4% | Not measured | Only Claude Opus 5 measured |
| BALROG | 63.4% | Not measured | Only Claude Opus 5 measured |
| PostTrainBench | 35.0% | Not measured | Only Claude Opus 5 measured |
| Vending-Bench 2 | $11,182 | $3,285 | Claude Opus 5 |
| LMArena Search | 1242 | Not measured | Only Claude Opus 5 measured |
| Neither model has a result on the other 12 boards here | |||
ReasoningLike-for-likePublic boards, shown as published, kept out of the index70
| Benchmark | Claude Opus 5 | DeepSeek V4 Pro 0423 | Leader |
|---|---|---|---|
| GPQA Diamond | 93.4% | 89.4% | Claude Opus 5 |
| LiveBench reasoning | 91.2% | Not measured | Only Claude Opus 5 measured |
| MysteryMechanism (Vals AI) | 37.4% | Not measured | Only Claude Opus 5 measured |
| MMMU Pro (Vals AI) | 89.9% | Not measured | Only Claude Opus 5 measured |
| GPQA Diamond (Artificial Analysis run) | 93.7% | Not measured | Only Claude Opus 5 measured |
| Humanity's Last Exam (Artificial Analysis run) | 54.9% | Not measured | Only Claude Opus 5 measured |
| GPQA Diamond (Epoch AI run) | 93.9% | 90.9% | Claude Opus 5 |
| ARC-AGI | 97.5% | Not measured | Only Claude Opus 5 measured |
| ARC-AGI-2 | 90.4% | Not measured | Only Claude Opus 5 measured |
| SimpleBench | 80.6% | Not measured | Only Claude Opus 5 measured |
| Chess puzzles | 43.0% | 19.0% | Claude Opus 5 |
| Mystery game puzzles | 59.0% | 17.0% | Claude Opus 5 |
| EBR-bench | 45.7% | Not measured | Only Claude Opus 5 measured |
| LMArena Text: hard prompts | 1527 | 1461 | Claude Opus 5 |
| LMArena Text: expert | 1554 | 1463 | Claude Opus 5 |
| LMArena Text: hard prompts in English | 1523 | 1466 | Claude Opus 5 |
| LMArena Vision: diagrams | 1348 | Not measured | Only Claude Opus 5 measured |
| LMArena Vision: homework | 1338 | Not measured | Only Claude Opus 5 measured |
| Neither model has a result on the other 3 boards here | |||
MathLike-for-likePublic boards, shown as published, kept out of the index60
| Benchmark | Claude Opus 5 | DeepSeek V4 Pro 0423 | Leader |
|---|---|---|---|
| OTIS Mock AIME | 98.9% | 96.7% | Claude Opus 5 |
| FrontierMath, tiers 1 to 3 | 85.6% | 45.3% | Claude Opus 5 |
| FrontierMath, tier 4 | 73.2% | 2.4% | Claude Opus 5 |
| ProofBench | 99.0% | 16.0% | Claude Opus 5 |
| LMArena Text: math | 1533 | 1443 | Claude Opus 5 |
| LMArena Text: mathematical work | 1544 | 1455 | Claude Opus 5 |
| Neither model has a result on the other 7 boards here | |||
KnowledgeLike-for-likePublic boards, shown as published, kept out of the index100
| Benchmark | Claude Opus 5 | DeepSeek V4 Pro 0423 | Leader |
|---|---|---|---|
| MMLU Pro | 91.6% | 87.3% | Claude Opus 5 |
| LegalBench (Vals AI) | 87.0% | 80.3% | Claude Opus 5 |
| CaseLaw (Vals AI) | Not measured | 59.4% | Only DeepSeek V4 Pro 0423 measured |
| CorpFin (Vals AI) | 73.2% | 61.4% | Claude Opus 5 |
| TaxEval (Vals AI) | 75.1% | 72.1% | Claude Opus 5 |
| MortgageTax (Vals AI) | 72.1% | Not measured | Only Claude Opus 5 measured |
| MedCode (Vals AI) | 63.6% | 40.5% | Claude Opus 5 |
| SimpleQA Verified | 59.9% | 47.0% | Claude Opus 5 |
| LMArena Text: science | 1531 | 1470 | Claude Opus 5 |
| LMArena Text: medicine and healthcare | 1525 | 1466 | Claude Opus 5 |
| LMArena Text: legal and government | 1530 | 1464 | Claude Opus 5 |
| LMArena Text: business and finance | 1501 | 1443 | Claude Opus 5 |
| Neither model has a result on the other 5 boards here | |||
Web designLike-for-likePublic boards, shown as published, kept out of the index80
| Benchmark | Claude Opus 5 | DeepSeek V4 Pro 0423 | Leader |
|---|---|---|---|
| Design Arena: websites | 1309 | 1235 | Claude Opus 5 |
| Design Arena: UI components | 1348 | 1221 | Claude Opus 5 |
| Design Arena: game dev | 1339 | 1229 | Claude Opus 5 |
| Design Arena: data viz | 1329 | 1203 | Claude Opus 5 |
| Design Arena: 3D | 1335 | 1254 | Claude Opus 5 |
| Design Arena: all code categories | 1324 | 1240 | Claude Opus 5 |
| Design Arena: SVG | 1310 | 1139 | Claude Opus 5 |
| Design Arena: ASCII art | 1352 | 1148 | Claude Opus 5 |
| LiveBench design | 83.0% | Not measured | Only Claude Opus 5 measured |
GeneralLike-for-likePublic boards, shown as published, kept out of the index210
| Benchmark | Claude Opus 5 | DeepSeek V4 Pro 0423 | Leader |
|---|---|---|---|
| LiveBench data analysis | 74.6% | Not measured | Only Claude Opus 5 measured |
| Public Benefits Bench (Vals AI) | 76.9% | 62.9% | Claude Opus 5 |
| SAGE (Vals AI) | 49.4% | Not measured | Only Claude Opus 5 measured |
| MedScribe (Vals AI) | 91.0% | 75.1% | Claude Opus 5 |
| Vals Multimodal Index | 73.9% | Not measured | Only Claude Opus 5 measured |
| Long context reasoning (AA-LCR) | 82.0% | Not measured | Only Claude Opus 5 measured |
| DTBench | 97.9% | 90.7% | Claude Opus 5 |
| LMCA | 64.5% | 41.2% | Claude Opus 5 |
| Surface Evolver Bench | Not measured | 40.0% | Only DeepSeek V4 Pro 0423 measured |
| Furniture assembly | 60.8% | Not measured | Only Claude Opus 5 measured |
| CL-bench Life | Not measured | 13.5% | Only DeepSeek V4 Pro 0423 measured |
| LMArena Text | 1509 | 1451 | Claude Opus 5 |
| LMArena Text: instruction following | 1518 | 1445 | Claude Opus 5 |
| LMArena Text: creative writing | 1493 | 1446 | Claude Opus 5 |
| LMArena Vision | 1320 | Not measured | Only Claude Opus 5 measured |
| LMArena Text: English | 1508 | 1459 | Claude Opus 5 |
| LMArena Text: non-English | 1502 | 1439 | Claude Opus 5 |
| LMArena Text: Chinese | 1573 | 1481 | Claude Opus 5 |
| LMArena Text: German | 1527 | 1458 | Claude Opus 5 |
| LMArena Text: Japanese | 1520 | 1446 | Claude Opus 5 |
| LMArena Text: Korean | 1517 | 1409 | Claude Opus 5 |
| LMArena Text: French | 1519 | 1473 | Claude Opus 5 |
| LMArena Text: Spanish | 1523 | 1457 | Claude Opus 5 |
| LMArena Text: Russian | 1507 | 1453 | Claude Opus 5 |
| LMArena Text: Polish | 1522 | 1461 | Claude Opus 5 |
| LMArena Text: multi-turn | 1500 | 1468 | Claude Opus 5 |
| LMArena Text: longer queries | 1517 | 1458 | Claude Opus 5 |
| LMArena Text: writing and language | 1499 | 1447 | Claude Opus 5 |
| LMArena Text: entertainment and media | 1481 | 1431 | Claude Opus 5 |
| LMArena Vision: English | 1315 | Not measured | Only Claude Opus 5 measured |
| LMArena Vision: Chinese | 1400 | Not measured | Only Claude Opus 5 measured |
| LMArena Vision: reading text | 1329 | Not measured | Only Claude Opus 5 measured |
| LMArena Vision: humor | 1293 | Not measured | Only Claude Opus 5 measured |
| LMArena Vision: creative writing | 1313 | Not measured | Only Claude Opus 5 measured |
| LMArena Document | 1516 | Not measured | Only Claude Opus 5 measured |
| Neither model has a result on the other 11 boards here | |||
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
| Benchmark | Claude Opus 5 | DeepSeek V4 Pro 0423 | 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 0423 costs less
Repository review
50K input + 3K output tokens
DeepSeek V4 Pro 0423 costs less
Agent loop
200K input + 10K output tokens
DeepSeek V4 Pro 0423 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 Opus 5 | DeepSeek V4 Pro 0423 |
|---|---|---|
| Maker | Anthropic | DeepSeek |
| Weights | No public weights listed | Open weights |
| Released | July 24, 2026 | April 24, 2026 |
| Context window | 1M | 1.05M |
| Input $/Mtok | $5.00 | $0.9553 |
| Output $/Mtok | $25.00 | $1.9105 |
| Coding index | 81.8 | 35.3 |
| Independent measurements behind it | 2 | 1 |
| Cost per solved task | $1.33 | $0.57 |
| Source | OpenRouter listing | OpenRouter listing |
Questions
Is Claude Opus 5 better than DeepSeek V4 Pro 0423 for coding?
Pick Claude Opus 5 for coding: it scores 81.8 on the Vibe Coding Index against 35.3 for DeepSeek V4 Pro 0423. Pick DeepSeek V4 Pro 0423 when output tokens cost 13.1x less, or when a solved task costs $0.57 against $1.33.
Which is cheaper, Claude Opus 5 or DeepSeek V4 Pro 0423?
Claude Opus 5 costs $5.00 per million input tokens and $25.00 per million output tokens; DeepSeek V4 Pro 0423 costs $0.9553 and $1.9105. Per solved SWE-bench Verified task, Claude Opus 5 costs $1.33 and DeepSeek V4 Pro 0423 $0.57.
Which has the bigger context window, Claude Opus 5 or DeepSeek V4 Pro 0423?
DeepSeek V4 Pro 0423, with 1.05M tokens against 1M.
How many benchmarks is the Claude Opus 5 vs DeepSeek V4 Pro 0423 comparison based on?
80 benchmarks have a result for both models as of October 21, 2026. Claude Opus 5's coding index rests on 2 independent benchmarks and DeepSeek V4 Pro 0423's on 1. Where only one model has a result, the page says so instead of naming a winner.