LLMBenchmark.ai

Enter opens the highlighted result. Esc closes.

Claude Sonnet 4.6 vs DeepSeek V4 Pro 0423

Updated October 21, 2026. 75 benchmarks have a result for both. Coding is ranked on the Vibe Coding Index; every other board is shown as published.

Replace Claude Sonnet 4.6 Replace DeepSeek V4 Pro 0423
Claude Sonnet 4.6Anthropic
Coding
57.8
Overall
30.1
Output $/M
$15.00
53
DeepSeek V4 Pro 0423DeepSeek
Coding
35.3
Overall
30.4
Output $/M
$1.9105
21
vs boards won of 75 shared, 1 tied

Decision reading

Pick Claude Sonnet 4.6 for coding: it scores 57.8 on the Vibe Coding Index against 35.3 for DeepSeek V4 Pro 0423.

Pick DeepSeek V4 Pro 0423 when output tokens cost 7.9x less, or when a solved task costs $0.57 against $1.68.

Model AClaude Sonnet 4.6Anthropic57.8/1002 independent benchmarks$1.68 per solved taskModel BDeepSeek V4 Pro 0423DeepSeek35.3/1001 independent benchmark$0.57 per solved task
Shared results
75
Claude Sonnet 4.6 only
42
DeepSeek V4 Pro 0423 only
8
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.

At a glance

Coding index
Claude Sonnet 4.6 57.8 from 2 benchmarks, DeepSeek V4 Pro 0423 35.3 from 1 benchmark.
Widest category
Agentic work, 15.5 points apart on average over 9 shared benchmarks.
Widest single gap
Design Arena: agents building full-stack apps: Claude Sonnet 4.6 1213 against DeepSeek V4 Pro 0423 948, 33.1 points apart on a 0 to 100 scale.
Output price
Claude Sonnet 4.6 $15.00 against DeepSeek V4 Pro 0423 $1.9105 per million output tokens, 7.9x apart.
Context window
Claude Sonnet 4.6 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.

Boards won, where both have a result
CategoryClaude Sonnet 4.6BarsDeepSeek V4 Pro 0423Ties
Coding1221 tie
Intelligence11
Agentic work90
Reasoning43
Math31
Knowledge82
Web design71
General911
All shared boards53211 tie
CodingLike-for-likeFour of these boards feed the Vibe Coding Index122
Coding benchmarks for Claude Sonnet 4.6 and DeepSeek V4 Pro 0423
BenchmarkClaude Sonnet 4.6DeepSeek V4 Pro 0423Leader
LiveBench coding66.6%Not measuredOnly Claude Sonnet 4.6 measured
LMArena WebDev15231447Claude Sonnet 4.6
AA Coding Index63.059.4Claude Sonnet 4.6
LiveCodeBench (Vals AI)82.1%87.5%DeepSeek V4 Pro 0423
SWE-bench Verified (Vals AI)77.4%77.4%Tie
Terminal-Bench 4 (Vals AI)Not measured11.1%Only DeepSeek V4 Pro 0423 measured
Vibe Code Bench (Vals AI)55.8%49.9%Claude Sonnet 4.6
Code migration (Vals AI)39.9%26.2%Claude Sonnet 4.6
ProgramBench (Vals AI)0.5%0.0%Claude Sonnet 4.6
Terminal-Bench 2.1 (Vals AI)57.3%50.2%Claude Sonnet 4.6
SkillsBench (Vals AI)49.0%51.3%DeepSeek V4 Pro 0423
SciCode50.1%Not measuredOnly Claude Sonnet 4.6 measured
Terminal-Bench Hard (Artificial Analysis run)53.0%Not measuredOnly Claude Sonnet 4.6 measured
Terminal-Bench53.4%Not measuredOnly Claude Sonnet 4.6 measured
FrontierCode24.3%17.6%Claude Sonnet 4.6
WeirdML66.1%48.9%Claude Sonnet 4.6
ALE-BenchNot measured1006.1Only DeepSeek V4 Pro 0423 measured
LMArena Text: coding15031470Claude Sonnet 4.6
LMArena Text: software and IT14911469Claude Sonnet 4.6
LMArena WebDev: React15121441Claude Sonnet 4.6
LMArena WebDev: HTML15271435Claude Sonnet 4.6
LMArena WebDev: from an image1529Not measuredOnly Claude Sonnet 4.6 measured
Neither model has a result on the other 16 boards here
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index11
Intelligence benchmarks for Claude Sonnet 4.6 and DeepSeek V4 Pro 0423
BenchmarkClaude Sonnet 4.6DeepSeek V4 Pro 0423Leader
AA Intelligence Index30.130.4DeepSeek V4 Pro 0423
Epoch Capabilities Index152.8149.1Claude Sonnet 4.6
Agentic workLike-for-likePublic boards, shown as published, kept out of the index90
Agentic work benchmarks for Claude Sonnet 4.6 and DeepSeek V4 Pro 0423
BenchmarkClaude Sonnet 4.6DeepSeek V4 Pro 0423Leader
AA Agentic Index31.826.3Claude Sonnet 4.6
Design Arena: agents building web apps11901000Claude Sonnet 4.6
Design Arena: agents building full-stack apps1213948Claude Sonnet 4.6
Design Arena: agents building mobile apps1212Not measuredOnly Claude Sonnet 4.6 measured
Design Arena: agents building Android apps1202Not measuredOnly Claude Sonnet 4.6 measured
Design Arena: agents building Godot games12261059Claude Sonnet 4.6
Design Arena: agents building games1177Not measuredOnly Claude Sonnet 4.6 measured
LiveBench research63.6%Not measuredOnly Claude Sonnet 4.6 measured
LiveBench browser68.0%Not measuredOnly Claude Sonnet 4.6 measured
LiveBench long agentic23.1%Not measuredOnly Claude Sonnet 4.6 measured
Finance Agent (Vals AI)51.0%44.1%Claude Sonnet 4.6
Tax Agent Bench (Vals AI)25.8%25.2%Claude Sonnet 4.6
Legal Research Bench (Vals AI)38.5%23.1%Claude Sonnet 4.6
Excel Modeling Benchmark (Vals AI)60.2%51.6%Claude Sonnet 4.6
τ²-Bench Telecom79.5%Not measuredOnly Claude Sonnet 4.6 measured
APEX-Agents43.0%Not measuredOnly Claude Sonnet 4.6 measured
OSWorld 2.09.3%Not measuredOnly Claude Sonnet 4.6 measured
DeepResearch Bench54.9%Not measuredOnly Claude Sonnet 4.6 measured
OSWorld72.1%Not measuredOnly Claude Sonnet 4.6 measured
Vending-Bench 2$7,204$3,285Claude Sonnet 4.6
Neither model has a result on the other 19 boards here
ReasoningLike-for-likePublic boards, shown as published, kept out of the index43
Reasoning benchmarks for Claude Sonnet 4.6 and DeepSeek V4 Pro 0423
BenchmarkClaude Sonnet 4.6DeepSeek V4 Pro 0423Leader
GPQA Diamond85.6%89.4%DeepSeek V4 Pro 0423
LiveBench reasoning86.4%Not measuredOnly Claude Sonnet 4.6 measured
MMMU Pro (Vals AI)83.6%Not measuredOnly Claude Sonnet 4.6 measured
GPQA Diamond (Artificial Analysis run)87.5%Not measuredOnly Claude Sonnet 4.6 measured
Humanity's Last Exam (Artificial Analysis run)33.6%Not measuredOnly Claude Sonnet 4.6 measured
GPQA Diamond (Epoch AI run)87.4%90.9%DeepSeek V4 Pro 0423
ARC-AGI86.5%Not measuredOnly Claude Sonnet 4.6 measured
ARC-AGI-260.4%Not measuredOnly Claude Sonnet 4.6 measured
Chess puzzles23.0%19.0%Claude Sonnet 4.6
Mystery game puzzles16.0%17.0%DeepSeek V4 Pro 0423
LMArena Text: hard prompts14851461Claude Sonnet 4.6
LMArena Text: expert15001463Claude Sonnet 4.6
LMArena Text: hard prompts in English14931466Claude Sonnet 4.6
LMArena Vision: diagrams1306Not measuredOnly Claude Sonnet 4.6 measured
LMArena Vision: homework1311Not measuredOnly Claude Sonnet 4.6 measured
Neither model has a result on the other 6 boards here
MathLike-for-likePublic boards, shown as published, kept out of the index31
Math benchmarks for Claude Sonnet 4.6 and DeepSeek V4 Pro 0423
BenchmarkClaude Sonnet 4.6DeepSeek V4 Pro 0423Leader
AIME92.3%Not measuredOnly Claude Sonnet 4.6 measured
OTIS Mock AIME85.8%96.7%DeepSeek V4 Pro 0423
FrontierMath, tiers 1 to 3Not measured45.3%Only DeepSeek V4 Pro 0423 measured
FrontierMath, tier 4Not measured2.4%Only DeepSeek V4 Pro 0423 measured
ProofBench45.0%16.0%Claude Sonnet 4.6
LMArena Text: math14621443Claude Sonnet 4.6
LMArena Text: mathematical work14831455Claude Sonnet 4.6
Neither model has a result on the other 6 boards here
KnowledgeLike-for-likePublic boards, shown as published, kept out of the index82
Knowledge benchmarks for Claude Sonnet 4.6 and DeepSeek V4 Pro 0423
BenchmarkClaude Sonnet 4.6DeepSeek V4 Pro 0423Leader
MMLU Pro87.3%87.3%Claude Sonnet 4.6
LegalBench (Vals AI)82.1%80.3%Claude Sonnet 4.6
CaseLaw (Vals AI)64.0%59.4%Claude Sonnet 4.6
CorpFin (Vals AI)65.3%61.4%Claude Sonnet 4.6
TaxEval (Vals AI)77.1%72.1%Claude Sonnet 4.6
MortgageTax (Vals AI)67.7%Not measuredOnly Claude Sonnet 4.6 measured
MedQA (Vals AI)92.1%Not measuredOnly Claude Sonnet 4.6 measured
MedCode (Vals AI)Not measured40.5%Only DeepSeek V4 Pro 0423 measured
SimpleQA Verified35.5%47.0%DeepSeek V4 Pro 0423
LMArena Text: science14741470Claude Sonnet 4.6
LMArena Text: medicine and healthcare14511466DeepSeek V4 Pro 0423
LMArena Text: legal and government14701464Claude Sonnet 4.6
LMArena Text: business and finance14641443Claude Sonnet 4.6
Neither model has a result on the other 4 boards here
Web designLike-for-likePublic boards, shown as published, kept out of the index71
Web design benchmarks for Claude Sonnet 4.6 and DeepSeek V4 Pro 0423
BenchmarkClaude Sonnet 4.6DeepSeek V4 Pro 0423Leader
Design Arena: websites12851235Claude Sonnet 4.6
Design Arena: UI components12711221Claude Sonnet 4.6
Design Arena: game dev12591229Claude Sonnet 4.6
Design Arena: data viz12781203Claude Sonnet 4.6
Design Arena: 3D12341254DeepSeek V4 Pro 0423
Design Arena: all code categories12761240Claude Sonnet 4.6
Design Arena: SVG11961139Claude Sonnet 4.6
Design Arena: ASCII art12351148Claude Sonnet 4.6
LiveBench design71.8%Not measuredOnly Claude Sonnet 4.6 measured
GeneralLike-for-likePublic boards, shown as published, kept out of the index911
General benchmarks for Claude Sonnet 4.6 and DeepSeek V4 Pro 0423
BenchmarkClaude Sonnet 4.6DeepSeek V4 Pro 0423Leader
LiveBench data analysis76.1%Not measuredOnly Claude Sonnet 4.6 measured
Public Benefits Bench (Vals AI)62.5%62.9%DeepSeek V4 Pro 0423
SAGE (Vals AI)46.6%Not measuredOnly Claude Sonnet 4.6 measured
MedScribe (Vals AI)Not measured75.1%Only DeepSeek V4 Pro 0423 measured
Vals Multimodal Index60.6%Not measuredOnly Claude Sonnet 4.6 measured
IFBench56.6%Not measuredOnly Claude Sonnet 4.6 measured
Long context reasoning (AA-LCR)80.0%Not measuredOnly Claude Sonnet 4.6 measured
DTBench89.9%90.7%DeepSeek V4 Pro 0423
LMCA46.5%41.2%Claude Sonnet 4.6
Surface Evolver BenchNot measured40.0%Only DeepSeek V4 Pro 0423 measured
ExploitBench23.6%Not measuredOnly Claude Sonnet 4.6 measured
CL-bench LifeNot measured13.5%Only DeepSeek V4 Pro 0423 measured
LMArena Text14581451Claude Sonnet 4.6
LMArena Text: instruction following14761445Claude Sonnet 4.6
LMArena Text: creative writing14351446DeepSeek V4 Pro 0423
LMArena Vision1283Not measuredOnly Claude Sonnet 4.6 measured
LMArena Text: English14721459Claude Sonnet 4.6
LMArena Text: non-English14401439Claude Sonnet 4.6
LMArena Text: Chinese14911481Claude Sonnet 4.6
LMArena Text: German14281458DeepSeek V4 Pro 0423
LMArena Text: Japanese14201446DeepSeek V4 Pro 0423
LMArena Text: Korean14101409Claude Sonnet 4.6
LMArena Text: French14651473DeepSeek V4 Pro 0423
LMArena Text: Spanish14641457Claude Sonnet 4.6
LMArena Text: Russian14401453DeepSeek V4 Pro 0423
LMArena Text: Polish14431461DeepSeek V4 Pro 0423
LMArena Text: multi-turn14641468DeepSeek V4 Pro 0423
LMArena Text: longer queries14791458Claude Sonnet 4.6
LMArena Text: writing and language14441447DeepSeek V4 Pro 0423
LMArena Text: entertainment and media14311431DeepSeek V4 Pro 0423
LMArena Vision: English1298Not measuredOnly Claude Sonnet 4.6 measured
LMArena Vision: Chinese1307Not measuredOnly Claude Sonnet 4.6 measured
LMArena Vision: reading text1296Not measuredOnly Claude Sonnet 4.6 measured
LMArena Vision: humor1248Not measuredOnly Claude Sonnet 4.6 measured
LMArena Vision: recognizing things1195Not measuredOnly Claude Sonnet 4.6 measured
LMArena Vision: creative writing1272Not measuredOnly Claude Sonnet 4.6 measured
LMArena Document1482Not measuredOnly Claude Sonnet 4.6 measured
Neither model has a result on the other 9 boards here
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
Image generation benchmarks for Claude Sonnet 4.6 and DeepSeek V4 Pro 0423
BenchmarkClaude Sonnet 4.6DeepSeek V4 Pro 0423Leader
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 Sonnet 4.6$0.011Fits in one request
DeepSeek V4 Pro 0423$0.0019Fits in one request

DeepSeek V4 Pro 0423 costs less

Repository review

50K input + 3K output tokens

Claude Sonnet 4.6$0.195Fits in one request
DeepSeek V4 Pro 0423$0.053Fits in one request

DeepSeek V4 Pro 0423 costs less

Agent loop

200K input + 10K output tokens

Claude Sonnet 4.6$0.750Fits in one request
DeepSeek V4 Pro 0423$0.210Fits in one request

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.

Specifications of Claude Sonnet 4.6 and DeepSeek V4 Pro 0423
SpecificationAnthropic Claude Sonnet 4.6DeepSeek V4 Pro 0423
MakerAnthropicDeepSeek
WeightsNo public weights listedOpen weights
ReleasedFebruary 17, 2026April 24, 2026
Context window1M1.05M
Input $/Mtok$3.00$0.9553
Output $/Mtok$15.00$1.9105
Coding index57.835.3
Independent measurements behind it21
Cost per solved task$1.68$0.57
SourceOpenRouter listingOpenRouter listing

Questions

Is Claude Sonnet 4.6 better than DeepSeek V4 Pro 0423 for coding?

Pick Claude Sonnet 4.6 for coding: it scores 57.8 on the Vibe Coding Index against 35.3 for DeepSeek V4 Pro 0423. Pick DeepSeek V4 Pro 0423 when output tokens cost 7.9x less, or when a solved task costs $0.57 against $1.68.

Which is cheaper, Claude Sonnet 4.6 or DeepSeek V4 Pro 0423?

Claude Sonnet 4.6 costs $3.00 per million input tokens and $15.00 per million output tokens; DeepSeek V4 Pro 0423 costs $0.9553 and $1.9105. Per solved SWE-bench Verified task, Claude Sonnet 4.6 costs $1.68 and DeepSeek V4 Pro 0423 $0.57.

Which has the bigger context window, Claude Sonnet 4.6 or DeepSeek V4 Pro 0423?

DeepSeek V4 Pro 0423, with 1.05M tokens against 1M.

How many benchmarks is the Claude Sonnet 4.6 vs DeepSeek V4 Pro 0423 comparison based on?

75 benchmarks have a result for both models as of October 21, 2026. Claude Sonnet 4.6'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.

Compare next