LLMBenchmark.ai

Enter opens the highlighted result. Esc closes.

Claude Opus 5 vs DeepSeek V4 Pro 0813

Updated October 21, 2026. 76 benchmarks have a result for both. Coding is ranked on the Vibe Coding Index; every other board is shown as published.

Replace Claude Opus 5 Replace DeepSeek V4 Pro 0813
Claude Opus 5Anthropic
Coding
81.8
Overall
50.8
Output $/M
$25.00
74
DeepSeek V4 Pro 0813DeepSeek
Coding
54.6
Overall
36.0
Output $/M
$1.98
2
vs boards won of 76 shared

Decision reading

Pick Claude Opus 5 for coding: it scores 81.8 on the Vibe Coding Index against 54.6 for DeepSeek V4 Pro 0813.

Pick DeepSeek V4 Pro 0813 when output tokens cost 12.6x less, or when a solved task costs $0.10 against $1.33.

Model AClaude Opus 5Anthropic81.8/1002 independent benchmarks$1.33 per solved taskModel BDeepSeek V4 Pro 0813DeepSeek54.6/1001 independent benchmark$0.10 per solved task
Shared results
76
Claude Opus 5 only
54
DeepSeek V4 Pro 0813 only
2
Like-for-like categories
7 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.

At a glance

Coding index
Claude Opus 5 81.8 from 2 benchmarks, DeepSeek V4 Pro 0813 54.6 from 1 benchmark.
Widest category
Math, 24.8 points apart on average over 6 shared benchmarks.
Widest single gap
ProofBench: Claude Opus 5 99.0% against DeepSeek V4 Pro 0813 50.0%, 49.0 points apart on a 0 to 100 scale.
Output price
Claude Opus 5 $25.00 against DeepSeek V4 Pro 0813 $1.98 per million output tokens, 12.6x apart.
Context window
Claude Opus 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.

Boards won, where both have a result
CategoryClaude Opus 5BarsDeepSeek V4 Pro 0813Ties
Coding201
Intelligence20
Agentic work80
Reasoning101
Math60
Knowledge100
General180
All shared boards742
CodingLike-for-likeFour of these boards feed the Vibe Coding Index201
Coding benchmarks for Claude Opus 5 and DeepSeek V4 Pro 0813
BenchmarkClaude Opus 5DeepSeek V4 Pro 0813Leader
LiveBench coding90.8%Not measuredOnly Claude Opus 5 measured
LMArena WebDev16891582Claude Opus 5
AA Coding Index78.068.8Claude Opus 5
LiveCodeBench (Vals AI)89.0%87.5%Claude Opus 5
SWE-bench Verified (Vals AI)97.0%96.4%Claude Opus 5
Terminal-Bench 4 (Vals AI)53.5%14.1%Claude Opus 5
Vibe Code Bench (Vals AI)88.4%82.3%Claude Opus 5
Code migration (Vals AI)57.5%41.5%Claude Opus 5
ProgramBench (Vals AI)3.0%0.0%Claude Opus 5
IOI (Vals AI)84.3%51.6%Claude Opus 5
Terminal-Bench 2.1 (Vals AI)84.6%54.7%Claude Opus 5
Vibe Code Bench 1-100 (Vals AI)28.5%17.5%Claude Opus 5
SkillsBench (Vals AI)60.4%53.8%Claude Opus 5
CyberBench (Vals AI)65.4%72.9%DeepSeek V4 Pro 0813
SciCode56.4%51.0%Claude Opus 5
FrontierCode53.4%28.5%Claude Opus 5
FrontierSWE52.0%Not measuredOnly Claude Opus 5 measured
WeirdML91.8%66.2%Claude Opus 5
ALE-Bench2164.61403.2Claude Opus 5
LMArena Text: coding15331469Claude Opus 5
LMArena Text: software and IT15271462Claude Opus 5
LMArena WebDev: React17111588Claude Opus 5
LMArena WebDev: HTML16801566Claude Opus 5
LMArena WebDev: from an image1662Not measuredOnly 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
Intelligence benchmarks for Claude Opus 5 and DeepSeek V4 Pro 0813
BenchmarkClaude Opus 5DeepSeek V4 Pro 0813Leader
AA Intelligence Index50.836.0Claude Opus 5
Epoch Capabilities Index163.9155.6Claude Opus 5
Agentic workLike-for-likePublic boards, shown as published, kept out of the index80
Agentic work benchmarks for Claude Opus 5 and DeepSeek V4 Pro 0813
BenchmarkClaude Opus 5DeepSeek V4 Pro 0813Leader
AA Agentic Index56.541.3Claude Opus 5
Design Arena: agents building web apps1266Not measuredOnly Claude Opus 5 measured
Design Arena: agents building full-stack apps1306Not measuredOnly Claude Opus 5 measured
Design Arena: agents building mobile apps1348Not measuredOnly Claude Opus 5 measured
Design Arena: agents building Android apps1258Not measuredOnly Claude Opus 5 measured
Design Arena: agents building Godot games1279Not measuredOnly Claude Opus 5 measured
Design Arena: agents building games1261Not measuredOnly Claude Opus 5 measured
Design Arena: PowerPoint slides1265Not measuredOnly Claude Opus 5 measured
LiveBench research88.9%Not measuredOnly Claude Opus 5 measured
LiveBench browser84.0%Not measuredOnly Claude Opus 5 measured
LiveBench long agentic62.1%Not measuredOnly Claude Opus 5 measured
SREBench (Vals AI)12.2%1.9%Claude Opus 5
Finance Agent (Vals AI)58.6%50.4%Claude Opus 5
Tax Agent Bench (Vals AI)45.5%26.8%Claude Opus 5
Legal Research Bench (Vals AI)55.3%40.9%Claude Opus 5
Harvey's Legal Agent Benchmark (Vals AI)Not measured10.8%Only DeepSeek V4 Pro 0813 measured
Excel Modeling Benchmark (Vals AI)73.6%52.8%Claude Opus 5
BioMysteryBench (Vals AI)79.3%Not measuredOnly Claude Opus 5 measured
Terminal-Bench-Science (Vals AI)27.1%4.3%Claude Opus 5
Vals Science Index47.9%Not measuredOnly Claude Opus 5 measured
Time Horizon Index (Vals AI)18.8%Not measuredOnly Claude Opus 5 measured
CUA-bench (Vals AI)9.0%Not measuredOnly Claude Opus 5 measured
APEX-Agents65.8%47.3%Claude Opus 5
OSWorld 2.031.4%Not measuredOnly Claude Opus 5 measured
BALROG63.4%Not measuredOnly Claude Opus 5 measured
PostTrainBench35.0%Not measuredOnly Claude Opus 5 measured
Vending-Bench 2$11,182Not measuredOnly Claude Opus 5 measured
LMArena Search1242Not measuredOnly Claude Opus 5 measured
Neither model has a result on the other 11 boards here
ReasoningLike-for-likePublic boards, shown as published, kept out of the index101
Reasoning benchmarks for Claude Opus 5 and DeepSeek V4 Pro 0813
BenchmarkClaude Opus 5DeepSeek V4 Pro 0813Leader
GPQA Diamond93.4%92.4%Claude Opus 5
LiveBench reasoning91.2%Not measuredOnly Claude Opus 5 measured
MysteryMechanism (Vals AI)37.4%Not measuredOnly Claude Opus 5 measured
MMMU Pro (Vals AI)89.9%Not measuredOnly Claude Opus 5 measured
GPQA Diamond (Artificial Analysis run)93.7%92.8%Claude Opus 5
Humanity's Last Exam (Artificial Analysis run)54.9%41.0%Claude Opus 5
GPQA Diamond (Epoch AI run)93.9%91.7%Claude Opus 5
ARC-AGI97.5%90.5%Claude Opus 5
ARC-AGI-290.4%61.3%Claude Opus 5
SimpleBench80.6%Not measuredOnly Claude Opus 5 measured
Chess puzzles43.0%46.0%DeepSeek V4 Pro 0813
Mystery game puzzles59.0%43.0%Claude Opus 5
EBR-bench45.7%Not measuredOnly Claude Opus 5 measured
LMArena Text: hard prompts15271458Claude Opus 5
LMArena Text: expert15541459Claude Opus 5
LMArena Text: hard prompts in English15231463Claude Opus 5
LMArena Vision: diagrams1348Not measuredOnly Claude Opus 5 measured
LMArena Vision: homework1338Not measuredOnly 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
Math benchmarks for Claude Opus 5 and DeepSeek V4 Pro 0813
BenchmarkClaude Opus 5DeepSeek V4 Pro 0813Leader
OTIS Mock AIME98.9%98.6%Claude Opus 5
FrontierMath, tiers 1 to 385.6%64.6%Claude Opus 5
FrontierMath, tier 473.2%26.8%Claude Opus 5
ProofBench99.0%50.0%Claude Opus 5
LMArena Text: math15331441Claude Opus 5
LMArena Text: mathematical work15441443Claude 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
Knowledge benchmarks for Claude Opus 5 and DeepSeek V4 Pro 0813
BenchmarkClaude Opus 5DeepSeek V4 Pro 0813Leader
MMLU Pro91.6%87.0%Claude Opus 5
LegalBench (Vals AI)87.0%82.4%Claude Opus 5
CorpFin (Vals AI)73.2%65.4%Claude Opus 5
TaxEval (Vals AI)75.1%73.1%Claude Opus 5
MortgageTax (Vals AI)72.1%Not measuredOnly Claude Opus 5 measured
MedCode (Vals AI)63.6%42.5%Claude Opus 5
SimpleQA Verified59.9%52.9%Claude Opus 5
LMArena Text: science15311457Claude Opus 5
LMArena Text: medicine and healthcare15251446Claude Opus 5
LMArena Text: legal and government15301459Claude Opus 5
LMArena Text: business and finance15011442Claude Opus 5
Neither model has a result on the other 6 boards here
Web designDirectional onlyPublic boards, shown as published, kept out of the index9 boards, none shared
Web design benchmarks for Claude Opus 5 and DeepSeek V4 Pro 0813
BenchmarkClaude Opus 5DeepSeek V4 Pro 0813Leader
Design Arena: websites1309Not measuredOnly Claude Opus 5 measured
Design Arena: UI components1348Not measuredOnly Claude Opus 5 measured
Design Arena: game dev1339Not measuredOnly Claude Opus 5 measured
Design Arena: data viz1329Not measuredOnly Claude Opus 5 measured
Design Arena: 3D1335Not measuredOnly Claude Opus 5 measured
Design Arena: all code categories1324Not measuredOnly Claude Opus 5 measured
Design Arena: SVG1310Not measuredOnly Claude Opus 5 measured
Design Arena: ASCII art1352Not measuredOnly Claude Opus 5 measured
LiveBench design83.0%Not measuredOnly Claude Opus 5 measured
GeneralLike-for-likePublic boards, shown as published, kept out of the index180
General benchmarks for Claude Opus 5 and DeepSeek V4 Pro 0813
BenchmarkClaude Opus 5DeepSeek V4 Pro 0813Leader
Vals IndexNot measured53.0%Only DeepSeek V4 Pro 0813 measured
LiveBench data analysis74.6%Not measuredOnly Claude Opus 5 measured
Public Benefits Bench (Vals AI)76.9%Not measuredOnly Claude Opus 5 measured
SAGE (Vals AI)49.4%Not measuredOnly Claude Opus 5 measured
MedScribe (Vals AI)91.0%80.2%Claude Opus 5
Vals Multimodal Index73.9%Not measuredOnly Claude Opus 5 measured
Long context reasoning (AA-LCR)82.0%80.3%Claude Opus 5
DTBench97.9%93.9%Claude Opus 5
LMCA64.5%45.5%Claude Opus 5
Furniture assembly60.8%Not measuredOnly Claude Opus 5 measured
LMArena Text15091446Claude Opus 5
LMArena Text: instruction following15181448Claude Opus 5
LMArena Text: creative writing14931439Claude Opus 5
LMArena Vision1320Not measuredOnly Claude Opus 5 measured
LMArena Text: English15081457Claude Opus 5
LMArena Text: non-English15021431Claude Opus 5
LMArena Text: Chinese15731481Claude Opus 5
LMArena Text: German1527Not measuredOnly Claude Opus 5 measured
LMArena Text: Japanese1520Not measuredOnly Claude Opus 5 measured
LMArena Text: Korean15171449Claude Opus 5
LMArena Text: French15191446Claude Opus 5
LMArena Text: Spanish15231435Claude Opus 5
LMArena Text: Russian15071448Claude Opus 5
LMArena Text: Polish1522Not measuredOnly Claude Opus 5 measured
LMArena Text: multi-turn15001454Claude Opus 5
LMArena Text: longer queries15171458Claude Opus 5
LMArena Text: writing and language14991443Claude Opus 5
LMArena Text: entertainment and media14811424Claude Opus 5
LMArena Vision: English1315Not measuredOnly Claude Opus 5 measured
LMArena Vision: Chinese1400Not measuredOnly Claude Opus 5 measured
LMArena Vision: reading text1329Not measuredOnly Claude Opus 5 measured
LMArena Vision: humor1293Not measuredOnly Claude Opus 5 measured
LMArena Vision: creative writing1313Not measuredOnly Claude Opus 5 measured
LMArena Document1516Not measuredOnly Claude Opus 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
Image generation benchmarks for Claude Opus 5 and DeepSeek V4 Pro 0813
BenchmarkClaude Opus 5DeepSeek V4 Pro 0813Leader
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

Claude Opus 5$0.018Fits in one request
DeepSeek V4 Pro 0813$0.0016Fits in one request

DeepSeek V4 Pro 0813 costs less

Repository review

50K input + 3K output tokens

Claude Opus 5$0.325Fits in one request
DeepSeek V4 Pro 0813$0.039Fits in one request

DeepSeek V4 Pro 0813 costs less

Agent loop

200K input + 10K output tokens

Claude Opus 5$1.25Fits in one request
DeepSeek V4 Pro 0813$0.152Fits in one request

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.

Specifications of Claude Opus 5 and DeepSeek V4 Pro 0813
SpecificationAnthropic Claude Opus 5DeepSeek V4 Pro 0813
MakerAnthropicDeepSeek
WeightsNo public weights listedOpen weights
ReleasedJuly 24, 2026August 12, 2026
Context window1M1.05M
Input $/Mtok$5.00$0.66
Output $/Mtok$25.00$1.98
Coding index81.854.6
Independent measurements behind it21
Cost per solved task$1.33$0.10
SourceOpenRouter listingOpenRouter listing

Questions

Is Claude Opus 5 better than DeepSeek V4 Pro 0813 for coding?

Pick Claude Opus 5 for coding: it scores 81.8 on the Vibe Coding Index against 54.6 for DeepSeek V4 Pro 0813. Pick DeepSeek V4 Pro 0813 when output tokens cost 12.6x less, or when a solved task costs $0.10 against $1.33.

Which is cheaper, Claude Opus 5 or DeepSeek V4 Pro 0813?

Claude Opus 5 costs $5.00 per million input tokens and $25.00 per million output tokens; DeepSeek V4 Pro 0813 costs $0.66 and $1.98. Per solved SWE-bench Verified task, Claude Opus 5 costs $1.33 and DeepSeek V4 Pro 0813 $0.10.

Which has the bigger context window, Claude Opus 5 or DeepSeek V4 Pro 0813?

DeepSeek V4 Pro 0813, with 1.05M tokens against 1M.

How many benchmarks is the Claude Opus 5 vs DeepSeek V4 Pro 0813 comparison based on?

76 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 0813's on 1. Where only one model has a result, the page says so instead of naming a winner.

Compare next