BenchLeader

Claude Sonnet 5 vs DeepSeek V4 Pro

Full profiles: Claude Sonnet 5 (Anthropic) and DeepSeek V4 Pro (DeepSeek): every benchmark, price and rank.

Updated · Data from 8 sources

As of 16 Nov 2026, Claude Sonnet 5 and DeepSeek V4 Pro are neck and neck on quality (59.0 vs 56.0 on the BenchLeader Index, inside the ±5.1 margin), and DeepSeek V4 Pro is 4.0× cheaper at $0.990 vs $4.00 per 1M tokens blended. Choose Claude Sonnet 5 for reasoning (65 vs 56) and coding (60 vs 52) and fast first answers (15.8 s vs 42.8 s), where it also streams ~2× faster; choose DeepSeek V4 Pro for long context (64 vs 63).

Key facts (as of 16 Nov 2026)

  • Claude Sonnet 5 scores 59.0 on the BenchLeader Index and DeepSeek V4 Pro scores 56.0 (BenchLeader, 16 Nov 2026).
  • Claude Sonnet 5 costs $4.00 per 1M tokens blended (list price, 16 Nov 2026).
  • DeepSeek V4 Pro costs $0.990 per 1M tokens blended (list price, 16 Nov 2026).
  • Claude Sonnet 5 streams ~2× faster than DeepSeek V4 Pro (Artificial Analysis, 7-day median, 16 Nov 2026).
  • Of the 21 benchmarks both models have, Claude Sonnet 5 leads 12 (BenchLeader, 16 Nov 2026).
  • On Arena WebDev, Claude Sonnet 5 is rated 1539 and DeepSeek V4 Pro 1581 (LMArena, as published by 16 Nov 2026).

Which to pick for…

The pick's numbers come first. Category scores on the BenchLeader scale (50 is the average model, 15 points one standard deviation). High confidence means a lead of half a standard deviation or more; under one point is called either.

Use casePickWhyConfidence
CodingClaude Sonnet 560 vs 52 category score; SciCode 54.3% vs 46.4%High
MathsEither65 vs 65 category score; Arena Maths 1472 vs 1465Either
ReasoningClaude Sonnet 565 vs 56 category score; Arena Hard Prompts 1490 vs 1482High
Long documentsDeepSeek V4 Pro64 vs 63 category score; AA-LCR v1.1 79.3% vs 76.7%Medium
Chat quality (human preference)Either65 vs 65 on Arena votesEither
Lowest costDeepSeek V4 Pro$1.98 vs $4.00 per 1M blended; agentic coding session $0.095 vs $0.160–
Fastest first answerClaude Sonnet 515.8 s (Claude Sonnet 5 (high)) vs 42.8 s (DeepSeek V4 Pro (high)) to the first answer token–
High volume on a budgetNeither flagshipA smaller model is usually the better fit: Cheapest good LLMs, Fastest LLMs–

Key differences

  • Claude Sonnet 5 (high) and DeepSeek V4 Pro (high) are neck and neck on quality (59.0 vs 56.0, within the ±5.1 margin).
  • Claude Sonnet 5 (high) is stronger in coding, maths, reasoning.
  • DeepSeek V4 Pro (high) is stronger in human preference, long context.
  • DeepSeek V4 Pro (high) is 4.0× cheaper ($0.990 vs $4.00 per 1M blended).

At a glance

SpecClaude Sonnet 5DeepSeek V4 Pro
DeveloperAnthropicDeepSeek
Released30 Jun 202613 Aug 2026
Context window (tokens)1,000,0001,048,576
Compared atClaude Sonnet 5 (high)DeepSeek V4 Pro (high)
Open weightsNoYes
Price in / out / cached input per 1M$2.00 / $10.00 / $0.200$1.32 / $3.96 / $0.022

Category by category

Category scores on the BenchLeader scale: 50 is the average evaluated model, 15 points is one standard deviation. A gap under one point is called level.

CategoryClaude Sonnet 5 (high)DeepSeek V4 Pro (high)Leader
Reasoning65.455.6Claude Sonnet 5
Coding59.951.5Claude Sonnet 5
Agents & tools52.9–not comparable
Maths65.364.5level
Knowledge–41.2not comparable
Human preference64.665.0level
Multimodal60.9–not comparable
Long context62.964.3DeepSeek V4 Pro

On one chart

Each domain on the index's own scale, where 50 is an average model. Click a domain to see the benchmarks behind it. Pick up to 5 models from the list.

2550 avg75ReasoningCodingMathsLong contextHumanpreferenceClaude Sonnet 5, Reasoning: 65.4Claude Sonnet 5, Coding: 59.9Claude Sonnet 5, Maths: 65.3Claude Sonnet 5, Long context: 62.9Claude Sonnet 5, Human preference: 64.6DeepSeek V4 Pro, Reasoning: 55.6DeepSeek V4 Pro, Coding: 51.5DeepSeek V4 Pro, Maths: 64.5DeepSeek V4 Pro, Long context: 64.3DeepSeek V4 Pro, Human preference: 65.0

Not drawn, measured for fewer than two of these models: Agents & tools, Knowledge, Instruction following, Multimodal.

Model, click to add or removeIndex

Where sources disagree

Across the 21 benchmarks both have, Claude Sonnet 5 leads 12, DeepSeek V4 Pro leads 9. Every overall index that rates both puts DeepSeek V4 Pro ahead.

Overall indexClaude Sonnet 5 (high)DeepSeek V4 Pro (high)Leader
BenchLeader Index59.056.0neck and neck
Arena Text14611464DeepSeek V4 Pro

Each publisher runs its own tasks, harnesses and effort settings, so their overall verdicts can differ; BenchLeader reconciles them on one scale. † measured at another effort setting of the same model (the publisher did not test the one compared here).

Price, speed and latency

List prices blended at three input tokens per output token. Per-task prices use the token counts on the methodology page. Cost, tokens and time per benchmark task are measured: Artificial Analysis’ averages over its whole Intelligence Index run.

MeasureClaude Sonnet 5 (high)DeepSeek V4 Pro (high)
Blended price per 1M tokens$4.00$1.98
Input / output per 1M$2.00 / $10.00$1.32 / $3.96
Cached input per 1M$0.200$0.022
Price per task: chat reply$0.0038$0.0017
Price per task: summarize a 30-page report$0.030$0.018
Price per task: code edit$0.027$0.014
Price per task: agentic coding session$0.160$0.095
Price per task: structured extraction$0.0060$0.0034
Agentic coding session, input mostly cached$0.079$0.037
Cost per benchmark task (AA Index)–$0.849
Output tokens per task (AA Index)–70k
Of which reasoning–45k
Time per task (AA Index)–15 min
Output speed99 tok/s80 tok/s
Time to first answer15.8 s42.8 s
Time to a complete answer, last measured 13 Nov 202640.7 s–

Each model is compared at its best-ranked reasoning-effort setting. Claude Sonnet 5 (high) scores 59.0 and takes 16 s to its first answer token; Claude Sonnet 5 (max) scores 56.1 in 158 s, Claude Sonnet 5 (xhigh) scores 55.4 in 32 s. DeepSeek V4 Pro (high) scores 56.0 and takes 43 s to its first answer token; DeepSeek V4 Pro (max) scores 54.7 in 47 s, DeepSeek V4 Pro scores 53.0 in 1 s. Higher effort means more thinking before the answer: better scores, longer waits and more tokens billed. How reasoning effort changes scores and cost.

Every benchmark

147 benchmarks, grouped by category; each name links to the board and its leaders. † measured at another effort setting of the same model, shown for reference and not counted as a win.

MetricPublisherClaude Sonnet 5 (high)DeepSeek V4 Pro (high)
BenchLeader IndexBenchLeader59.056.0
Agents & tools scoreBenchLeader52.9–
Coding scoreBenchLeader59.951.5
Human preference scoreBenchLeader64.665.0
Long context scoreBenchLeader62.964.3
Maths scoreBenchLeader65.364.5
Multimodal scoreBenchLeader60.9–
Reasoning scoreBenchLeader65.455.6
Knowledge scoreBenchLeader–41.2
Blended price $/MBenchLeader$4.00$1.98
Output speedBenchLeader99 tok/s80 tok/s
Time to first answerBenchLeader15.8 s42.8 s
Context windowBenchLeader1,000,000 tokens1,048,576 tokens
Reasoning
GPQA DiamondEpoch AI Benchmarking Hub80.3%†90.9%
SimpleBenchSimpleBench60.6%†–
Arena Hard PromptsArena (LMArena)14901482
LiveBench 1.0 ReasoningLiveBench88.7%†85.8%†
GPQA Diamond (AA)Artificial Analysis91.1%†92.8%†
Humanity's Last Exam (AA)Artificial Analysis35.7%41.5%
HLE-Diamond (AA)Artificial Analysis–20.5%
GPQA Diamond (Vals)Vals AI88.9%†92.4%†
Kagi LLM BenchmarkKagi LLM Benchmark–53.5%†
ARC-AGI-1ARC Prize–87.2%
ARC-AGI-2ARC Prize–59.7%
CritPtArtificial Analysis15.1%18.6%
CritPt v2 hard subset (AA)Artificial Analysis–28.5%
HLE-Diamond (Scale)Scale AI SEAL–13.4%†
Arena ExpertArena (LMArena)15091481
Chess PuzzlesEpoch AI Benchmarking Hub45.0%†16.0%
Mystery Game PuzzlesEpoch AI Benchmarking Hub16.0%†43.0%†
CL-bench LifeEpoch AI Benchmarking Hub–13.5%
LMCAEpoch AI Benchmarking Hub50.0%45.5%†
DTBenchEpoch AI Benchmarking Hub84.5%93.9%†
ForecastBenchForecastBench63.1%†60.7%†
BTF-3Epoch AI Benchmarking Hub14.2%†–
Coding
SWE-bench Verified (Epoch)Epoch AI Benchmarking Hub–77.6%†
SciCodeSciCode54.3%46.4%
WeirdMLWeirdML68.8%46.5%
Show all 147 benchmarks
MetricPublisherClaude Sonnet 5 (high)DeepSeek V4 Pro (high)
Coding
FrontierCodeCognition42.7%†28.6%
GSO-BenchGSO-Bench37.3%†–
Arena CodingArena (LMArena)15201503
Arena WebDevArena (LMArena)15391581
LiveBench 1.0 CodingLiveBench80.7%†77.2%†
SciCode (AA)Artificial Analysis54.3%51.7%
LiveCodeBenchVals AI82.4%†87.5%†
IOIVals AI45.0%†51.6%†
SWE-bench (Vals)Vals AI79.6%†96.4%†
Code MigrationVals AI44.4%†41.5%†
Code Migration v1.1Vals AI–52.0%†
ProgramBenchVals AI0.0%†0.0%†
Vibe Code Bench 1-100Vals AI13.8%†17.5%†
Vibe Code Bench v1.1Vals AI81.3%†82.3%†
SWE Atlas v1.0: Codebase QnAScale AI SEAL–27.1%†
SWE Atlas v1.0: Test WritingScale AI SEAL–27.1%†
SWE-Bench Pro V2Scale AI SEAL93.2%†–
SWE-Bench Pro V2 HARDScale AI SEAL88.2%†–
MirrorCodeEpoch AI Benchmarking Hub63.3%†33.3%†
Surface Evolver BenchEpoch AI Benchmarking Hub60.0%†40.0%
DeepSWE v1.1Epoch AI Benchmarking Hub48.2%–
CursorBenchCursor30.8%–
ALE-BenchEpoch AI Benchmarking Hub1463.11006.1
Terminal-Bench 4.0 (AA)Artificial Analysis5.0%6.1%
Terminal-Bench 2.1 (AA)Artificial Analysis80.5%†78.7%†
Startrise FrontendStartrise66.6%†62.7%†
GameBench 2 OverallDuelLab GameBench 246.6†41.9†
GameBench 2 by reasoning settingDuelLab GameBench 254.4†45.1
Design Arena WebsiteDesign Arena1268†1227†
FrontierCode 1.1 (AA)Artificial Analysis–38.8%†
FrontierCode 2.0 (AA)Artificial Analysis17.8%†16.7%
Agents & tools
Terminal-BenchTerminal-Bench12.1%†6.7%†
APEX-AgentsMercor54.5%†47.3%†
Arena AgentArena (LMArena)2–
LiveBench Agentic CodingLiveBench59.4%†55.0%†
Terminal-Bench 2.1 (Vals)Vals AI74.5%†50.2%†
HiL-BenchScale AI SEAL–38.5%†
GDPval-AA v2.1, normalised scoreArtificial Analysis–47.5%†
GDPval-AA v2.1Artificial Analysis12621336
τ³-Banking (AA)Artificial Analysis37.3%†39.6%†
Analyst Agent (AA)Artificial Analysis46.3%†–
CyberBenchVals AI61.9%†72.9%†
Finance Agent v2Vals AI53.9%†50.4%†
Finance Agent v2.1Vals AI–57.1%†
Harvey's Legal Agent Benchmark v1.1Vals AI–10.8%†
Harvey's Legal Agent Benchmark v1Vals AI5.0%†7.5%†
Legal Research BenchVals AI41.8%†40.9%†
SkillsBenchVals AI46.5%†53.8%†
SREBenchVals AI0.4%†1.9%†
Tax Agent BenchVals AI29.4%†26.8%†
Terminal-Bench 2.0 (Vals)Vals AI–56.2%†
Terminal-Bench 4.0 (Vals)Vals AI10.6%†14.1%†
Terminal-Bench ScienceVals AI5.7%†4.3%†
Arena SearchArena (LMArena)1191†–
Arena Agent Task CompletionArena (LMArena)-0.6–
GBAEvalEpoch AI Benchmarking Hub65.3%†–
Vending-Bench 2Andon Labs6377.7†3284.5†
AutomationBenchArtificial Analysis32.1%46.2%
GDP.pdf (AA)Artificial Analysis9.4%10.4%
EnterpriseOps-GymArtificial Analysis44.7%†49.6%†
AA-Briefcase v1.1Artificial Analysis11771148
Design Arena AgenticDesign Arena1222†927†
Terminal-Bench-Science 0.1 (AA)Artificial Analysis–1.9%
Maths
FrontierMath Tiers 1–3Epoch AI Benchmarking Hub65.6%†64.6%†
FrontierMath Tier 4Epoch AI Benchmarking Hub29.3%†26.8%†
OTIS Mock AIMEEpoch AI Benchmarking Hub80.0%†95.6%
ProofBenchVals AI77.0%†50.0%†
LiveBench MathematicsLiveBench92.9%†95.1%†
AIME 2026MathArena–96.7%†
HMMT February 2026MathArena–93.9%†
USAMO 2026MathArena–60.7%†
MathArena ApexMathArena–28.1%†
Arena MathsArena (LMArena)14721465
Knowledge
SimpleQA VerifiedEpoch AI Benchmarking Hub33.7%†52.9%†
LiveBench 1.0 Data AnalysisLiveBench71.7%†79.2%†
AA-Omniscience v1.1Artificial Analysis–-6.5
MMLU-ProVals AI87.5%†87.3%†
LegalBenchVals AI83.9%†82.4%†
CorpFinVals AI67.0%†65.4%†
TaxEvalVals AI75.6%†73.1%†
CaseLaw v2Vals AI–59.4%†
Excel Modeling BenchmarkVals AI66.3%†52.8%†
Excel Modeling Benchmark v1.1Vals AI–65.1%†
MedCodeVals AI47.5%†42.5%†
MedScribeVals AI76.0%†80.2%†
Public Benefits Bench v1.1Vals AI66.0%†62.9%†
Immigration BenchVals AI–61.7%†
DrugDiscoveryBenchScale AI SEAL50.0%†31.7%†
AA-Omniscience v1.1: accuracyArtificial Analysis–45.1%
AA-Omniscience v1.1: non-hallucinationArtificial Analysis–5.9%
FACTS ParametricGoogle20.5%†–
Instruction following
LiveBench LanguageLiveBench75.0%†82.1%†
LiveBench Instruction FollowingLiveBench63.9%†67.7%†
Arena Instruction FollowingArena (LMArena)14671461
Human preference
Arena TextArena (LMArena)14611464
EQ-Bench 4EQ-Bench1221†1156†
EQ-Bench Creative Writing v3EQ-Bench1798†1553†
Arena Creative WritingArena (LMArena)14341444
Arena Multi-turnArena (LMArena)14731469
Arena Longer QueriesArena (LMArena)14811474
Multimodal
Arena VisionArena (LMArena)1272–
MMMU-ProArtificial Analysis77.3%†–
MMMU-Pro (Vals)Vals AI83.0%†–
MortgageTaxVals AI70.0%†–
Humanity's Sixth SenseScale AI SEAL24.8%†–
GeoBenchGeoBench74.0%–
Blueprint-Bench 2Epoch AI Benchmarking Hub24.9%†–
Long context
AA-LCR v1.1Artificial Analysis76.7%79.3%
Arena DocumentArena (LMArena)1469–
MLCRArtificial Analysis4.4%–
Composite
Epoch Capabilities IndexEpoch AI Benchmarking Hub156.9†155.9†
LiveBench 1.0LiveBench76.0%†77.4%†
AA Intelligence Index v5.0.1Artificial Analysis–20.8
AA Intelligence Index v4.3.2Artificial Analysis31.736†
AA Progress Index v1Artificial Analysis296.8311.5†
Vals IndexVals AI–53.0†
Vals Finance IndexVals AI–52.9%†
Vals Multimodal IndexVals AI68.8%†–
Vals Router IndexVals AI–53.0%†
Vals Legal IndexVals AI–69.0%†
Safety & honesty
DistressBenchScale AI SEAL68.0%†–
Metadata
AA Openness IndexArtificial Analysis–44.4

Data as of 2026-11-16. Best configuration of each model; every score links to its source on the model pages.

DuelLab GameBench 2, release gb2-20261111-7b598bb13de4 (https://benchmarks.duellab.org/releases/gb2-20261111-7b598bb13de4/), CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Scores mapped to BenchLeader model names; not endorsed by DuelLab.

What changed

  • Claude Sonnet 5 (high): 2 new results (GDPval-AA v2.1, AA-Briefcase v1.1).
  • DeepSeek V4 Pro (high): BenchLeader Index 59.1 → 56.0.
  • DeepSeek V4 Pro (high): 17 new results (AA Intelligence Index v5.0.1, Humanity's Last Exam (AA), HLE-Diamond (AA)…); input price moved from $0.66/M to $1.32/M (100% up); output price moved from $1.98/M to $3.96/M (100% up).
  • Claude Sonnet 5 (high): BenchLeader Index 59.5 → 59.0.
  • Claude Sonnet 5 (high): 1 new result (GeoBench).
  • DeepSeek V4 Pro (high): BenchLeader Index 57.5 → 59.1.
  • Claude Sonnet 5 (high): 1 score revision.
  • DeepSeek V4 Pro (high): 1 score revision.
  • DeepSeek V4 Pro (high): 1 score revision.
  • Claude Sonnet 5 (high): 1 score revision.

Rankings: Best LLM overall · Best LLM for agents · Best cheap LLMs

Claude Sonnet 5 vs DeepSeek V4 Pro: questions

Is Claude Sonnet 5 better than DeepSeek V4 Pro?
Claude Sonnet 5 (high) and DeepSeek V4 Pro (high) are neck and neck on quality (59.0 vs 56.0, within the ±5.1 margin). The BenchLeader Index combines every independent quality benchmark; the two are within the margin of error as of 16 Nov 2026, so check the category scores for your use.
Claude Sonnet 5 vs DeepSeek V4 Pro: which is cheaper, and will that change?
DeepSeek V4 Pro is cheaper today: $0.990 against $4.00 per million tokens, blended at three input tokens per output token. No price change has been announced for either.
Is Claude Sonnet 5 or DeepSeek V4 Pro better for coding?
Claude Sonnet 5 scores higher in coding (60 vs 52 on the category index, where 50 is average). On Arena WebDev, the top shared coding benchmark, it is 1539 for Claude Sonnet 5 against 1581 for DeepSeek V4 Pro.
Which is faster, Claude Sonnet 5 or DeepSeek V4 Pro?
Claude Sonnet 5 streams ~2× faster (99 vs 80 output tokens per second, Artificial Analysis 7-day median).
Which is better for long documents?
DeepSeek V4 Pro scores higher on long-context benchmarks (64 vs 63). Both take about 1M tokens: Claude Sonnet 5 1,000,000, DeepSeek V4 Pro 1,048,576.
More questions (3)
Which has lower latency?
Claude Sonnet 5 answers first: 15.8 s against 42.8 s to the first answer token, as Artificial Analysis measures it. Reasoning effort drives most of this: a higher setting thinks longer before it answers.
Do independent benchmarks agree on Claude Sonnet 5 vs DeepSeek V4 Pro?
Across the 21 benchmarks both have, Claude Sonnet 5 leads 12, DeepSeek V4 Pro leads 9. Every overall index that rates both puts DeepSeek V4 Pro ahead.
When was DeepSeek V4 Pro released?
DeepSeek V4 Pro was released on 13 Aug 2026.