BenchLeader
Menu
Best LLMsChoose a modelCoding plansLeaderboardsModelsBenchmarksProvidersReleases12

DeepSeek V4 Pro vs Gemini 4 Carbon

Full profiles: DeepSeek V4 Pro (DeepSeek) and Gemini 4 Carbon (Google): every benchmark, price and rank.

Updated · Data from 9 sources

As of 9 Nov 2026, Gemini 4 Carbon is ahead of DeepSeek V4 Pro on quality (68.5 vs 59.1 on the BenchLeader Index), but DeepSeek V4 Pro is 8.1× cheaper at $0.990 vs $8.00 per 1M tokens blended. Gemini 4 Carbon leads in every category we measure both in. Gemini 4 Carbon streams ~1.5× faster.

Key facts (as of 9 Nov 2026)

  • DeepSeek V4 Pro scores 59.1 on the BenchLeader Index and Gemini 4 Carbon scores 68.5 (BenchLeader, 9 Nov 2026).
  • DeepSeek V4 Pro costs $0.990 per 1M tokens blended (list price, 9 Nov 2026).
  • Gemini 4 Carbon costs $8.00 per 1M tokens blended (list price, 9 Nov 2026).
  • Gemini 4 Carbon streams ~1.5× faster than DeepSeek V4 Pro (Artificial Analysis, 7-day median, 9 Nov 2026).
  • Of the 8 benchmarks both models have, Gemini 4 Carbon leads 8 (BenchLeader, 9 Nov 2026).
  • On Arena WebDev, DeepSeek V4 Pro is rated 1581 and Gemini 4 Carbon 1768 (LMArena, as published by 9 Nov 2026).
DeepSeek V4 Pro benchmarks high DeepSeek Gemini 4 Carbon benchmarks high GoogleAdd a third model

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
CodingGemini 4 Carbon75 vs 54 category score; Arena WebDev 1768 vs 1581High
ReasoningGemini 4 Carbon77 vs 62 category score; Arena Hard Prompts 1547 vs 1482High
Chat quality (human preference)Gemini 4 Carbon73 vs 65 on Arena votesHigh
Lowest costDeepSeek V4 Pro$0.990 vs $8.00 per 1M blended; agentic coding session $0.048 vs $0.320–
Fastest first answerDeepSeek V4 Pro1.5 s (DeepSeek V4 Pro (high)) vs 69.4 s (Gemini 4 Carbon (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

  • Gemini 4 Carbon (high) leads on quality: 68.5 vs 59.1.
  • DeepSeek V4 Pro (high) is stronger in maths.
  • Gemini 4 Carbon (high) is stronger in coding, human preference, reasoning, agents & tools, knowledge, long context.
  • DeepSeek V4 Pro (high) is 8.1× cheaper ($0.990 vs $8.00 per 1M blended).
  • Gemini 4 Carbon (high) streams 1.7× faster (75 vs 44 tokens per second).

At a glance

SpecDeepSeek V4 ProGemini 4 Carbon
DeveloperDeepSeekGoogle
Released13 Aug 202628 Oct 2026
Context window (tokens)1,048,5761,000,000
Compared atDeepSeek V4 Pro (high)Gemini 4 Carbon (high)
Open weightsYesNo
Price in / out / cached input per 1M$0.660 / $1.98 / $0.022$4.00 / $20.00 / –

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.

CategoryDeepSeek V4 Pro (high)Gemini 4 Carbon (high)Leader
Reasoning62.276.9Gemini 4 Carbon
Coding53.874.5Gemini 4 Carbon
Agents & tools–78.2not comparable
Maths64.6–not comparable
Knowledge–74.3not comparable
Human preference65.172.8Gemini 4 Carbon
Long context–64.3not comparable

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 avg75ReasoningCodingHumanpreferenceDeepSeek V4 Pro, Reasoning: 62.2DeepSeek V4 Pro, Coding: 53.8DeepSeek V4 Pro, Human preference: 65.1Gemini 4 Carbon, Reasoning: 76.9Gemini 4 Carbon, Coding: 74.5Gemini 4 Carbon, Human preference: 72.8

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

ModelIndex

Where sources disagree

Across the 8 benchmarks both have, DeepSeek V4 Pro leads 0, Gemini 4 Carbon leads 8. Every overall index that rates both puts Gemini 4 Carbon ahead.

Overall indexDeepSeek V4 Pro (high)Gemini 4 Carbon (high)Leader
BenchLeader Index59.168.5Gemini 4 Carbon
AA Intelligence Index v5.012.6†60.0Gemini 4 Carbon
Vals Index53.0†70.5Gemini 4 Carbon
Arena Text14641527Gemini 4 Carbon

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.

MeasureDeepSeek V4 Pro (high)Gemini 4 Carbon (high)
Blended price per 1M tokens$0.990$8.00
Input / output per 1M$0.660 / $1.98$4.00 / $20.00
Cached input per 1M$0.022–
Price per task: chat reply$0.0009$0.0076
Price per task: summarize a 30-page report$0.0091$0.060
Price per task: code edit$0.0069$0.054
Price per task: agentic coding session$0.048$0.320
Price per task: structured extraction$0.0017$0.012
Agentic coding session, input mostly cached$0.019–
Cost per benchmark task (AA Index)–$7.76
Output tokens per task (AA Index)–123k
Of which reasoning–72k
Time per task (AA Index)–27 min
Output speed44 tok/s75 tok/s
Time to first answer1.5 s69.4 s
Time to a complete answer–160.3 s

Each model is compared at its best-ranked reasoning-effort setting. DeepSeek V4 Pro (high) scores 59.1 and takes 1 s to its first answer token; DeepSeek V4 Pro (max) scores 56.9 in 74 s, DeepSeek V4 Pro scores 53.3 in 1 s. Gemini 4 Carbon (high) scores 68.5 and takes 69 s to its first answer token; Gemini 4 Carbon (medium) scores 64.1 in 27 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

121 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.

MetricPublisherDeepSeek V4 Pro (high)Gemini 4 Carbon (high)
BenchLeader IndexBenchLeader59.168.5
Coding scoreBenchLeader53.874.5
Human preference scoreBenchLeader65.172.8
Maths scoreBenchLeader64.6–
Reasoning scoreBenchLeader62.276.9
Agents & tools scoreBenchLeader–78.2
Knowledge scoreBenchLeader–74.3
Long context scoreBenchLeader–64.3
Blended price $/MBenchLeader$0.990$8.00
Output speedBenchLeader44 tok/s75 tok/s
Time to first answerBenchLeader1.5 s69.4 s
Context windowBenchLeader1,048,576 tokens1,000,000 tokens
Reasoning
GPQA DiamondEpoch AI Benchmarking Hub90.9%–
Arena Hard PromptsArena (LMArena)14821547
LiveBench 1.0 ReasoningLiveBench85.8%†–
GPQA Diamond (AA)Artificial Analysis92.8%†–
Humanity's Last Exam (AA)Artificial Analysis41.0%†–
HLE-Diamond (AA)Artificial Analysis19.5%†69.1%
GPQA Diamond (Vals)Vals AI92.4%†–
Kagi LLM BenchmarkKagi LLM Benchmark53.5%†–
ARC-AGI-1ARC Prize87.2%–
ARC-AGI-2ARC Prize59.7%–
CritPtArtificial Analysis18.0%†–
CritPt v2 hard subset (AA)Artificial Analysis23.5%†69.0%
Chess PuzzlesEpoch AI Benchmarking Hub16.0%–
Mystery Game PuzzlesEpoch AI Benchmarking Hub43.0%†–
CL-bench LifeEpoch AI Benchmarking Hub13.5%–
LMCAEpoch AI Benchmarking Hub45.5%†–
DTBenchEpoch AI Benchmarking Hub93.9%†–
ForecastBenchForecastBench60.7%†–
Coding
SWE-bench Verified (Epoch)Epoch AI Benchmarking Hub77.6%†–
SciCodeSciCode46.4%64.2%
WeirdMLWeirdML46.5%–
FrontierCodeCognition28.6%–
Arena CodingArena (LMArena)15031567
Arena WebDevArena (LMArena)15811768
LiveBench 1.0 CodingLiveBench77.2%†–
Show all 121 benchmarks
MetricPublisherDeepSeek V4 Pro (high)Gemini 4 Carbon (high)
Coding
SciCode (AA)Artificial Analysis51.0%†–
LiveCodeBenchVals AI87.5%†–
IOIVals AI51.6%†–
SWE-bench (Vals)Vals AI96.4%†–
Code MigrationVals AI41.5%†–
Code Migration v1.1Vals AI52.0%†–
ProgramBenchVals AI0.0%†–
Vibe Code Bench 1-100Vals AI17.5%†–
Vibe Code Bench v1.1Vals AI82.3%†89.1%
SWE Atlas: Codebase QnAScale AI SEAL27.1%†–
SWE Atlas: Test WritingScale AI SEAL27.1%†–
Surface Evolver BenchEpoch AI Benchmarking Hub40.0%–
ALE-BenchEpoch AI Benchmarking Hub1006.1–
Terminal-Bench 4.0 (AA)Artificial Analysis14.1%†65.2%
Terminal-Bench 2.1 (AA)Artificial Analysis78.7%†–
Startrise FrontendStartrise62.7%†–
GameBench 2 OverallDuelLab GameBench 242.0†–
GameBench 2 by reasoning settingDuelLab GameBench 245.0–
Design Arena WebsiteDesign Arena1231†–
FrontierCode 1.1 (AA)Artificial Analysis38.8%†–
FrontierCode 2.0 (AA)Artificial Analysis17.1%†40.6%
Agents & tools
APEX-AgentsMercor47.3%†–
Arena AgentArena (LMArena)–16.6
LiveBench Agentic CodingLiveBench55.0%†–
Terminal-Bench 2.1 (Vals)Vals AI50.2%†–
HiL-BenchScale AI SEAL38.5%†–
GDPval-AA v2.1, normalised scoreArtificial Analysis47.5%†–
GDPval-AA v2.1Artificial Analysis1457†1733
τ³-Banking (AA)Artificial Analysis39.6%†–
CyberBenchVals AI72.9%†–
Finance Agent v2Vals AI50.4%†–
Finance Agent v2.1Vals AI57.1%†67.2%
Harvey's Legal Agent Benchmark v1.1Vals AI10.8%†17.5%
Harvey's Legal Agent Benchmark v1Vals AI7.5%†–
Legal Research BenchVals AI40.9%†55.8%
SkillsBenchVals AI53.8%†–
SREBenchVals AI1.9%†–
Tax Agent BenchVals AI26.8%†51.8%
Terminal-Bench 2.0 (Vals)Vals AI56.2%†–
Terminal-Bench 4.0 (Vals)Vals AI14.1%†63.1%
Terminal-Bench ScienceVals AI4.3%†–
Vending-Bench 2Andon Labs3284.5†16484.2†
AutomationBenchArtificial Analysis56.7%†77.8%
GDP.pdf (AA)Artificial Analysis11.4%†28.4%
EnterpriseOps-GymArtificial Analysis49.6%†–
AA-Briefcase v1.1Artificial Analysis1253†1611
GDPxGDPx by SemiAnalysis–10.9%
Design Arena AgenticDesign Arena921†–
Terminal-Bench-Science 0.1 (AA)Artificial Analysis5.7%†64.8%
FACTS SearchGoogle–94.4%†
Maths
FrontierMath Tiers 1–3Epoch AI Benchmarking Hub64.6%†–
FrontierMath Tier 4Epoch AI Benchmarking Hub26.8%†–
OTIS Mock AIMEEpoch AI Benchmarking Hub95.6%–
ProofBenchVals AI50.0%†–
LiveBench MathematicsLiveBench95.1%†–
AIME 2026MathArena96.7%†–
HMMT February 2026MathArena93.9%†–
USAMO 2026MathArena60.7%†–
MathArena ApexMathArena28.1%†–
Arena MathsArena (LMArena)1465–
Knowledge
SimpleQA VerifiedEpoch AI Benchmarking Hub52.9%†–
LiveBench 1.0 Data AnalysisLiveBench79.2%†–
AA-Omniscience v1.1Artificial Analysis-6.4†53.5
MMLU-ProVals AI87.3%†–
LegalBenchVals AI82.4%†–
CorpFinVals AI65.4%†–
TaxEvalVals AI73.1%†–
CaseLaw v2Vals AI59.4%†–
Excel Modeling BenchmarkVals AI52.8%†–
Excel Modeling Benchmark v1.1Vals AI65.1%†76.5%
MedCodeVals AI42.5%†–
MedScribeVals AI80.2%†–
Public Benefits Bench v1.1Vals AI62.9%†–
DrugDiscoveryBenchScale AI SEAL31.7%†–
AA-Omniscience v1.1: accuracyArtificial Analysis45.2%†58.3%
AA-Omniscience v1.1: non-hallucinationArtificial Analysis5.9%†88.6%
Instruction following
LiveBench LanguageLiveBench82.1%†–
LiveBench Instruction FollowingLiveBench67.7%†–
Arena Instruction FollowingArena (LMArena)14611526
Human preference
Arena TextArena (LMArena)14641527
EQ-Bench 4EQ-Bench1156†–
Arena Creative WritingArena (LMArena)14441518
Arena Multi-turnArena (LMArena)14691564
Arena Longer QueriesArena (LMArena)14741539
Multimodal
MMMU-ProArtificial Analysis–89.3%†
Blueprint-Bench 2Epoch AI Benchmarking Hub–58.6%†
Long context
AA-LCR v1.1Artificial Analysis80.3%†79.3%
Composite
Epoch Capabilities IndexEpoch AI Benchmarking Hub155.9†–
LiveBench 1.0LiveBench77.4%†–
AA Intelligence Index v5.0Artificial Analysis23.1†60.0
AA Intelligence Index v4.3.2Artificial Analysis36†–
AA Progress Index v1Artificial Analysis311.5†350.8
Vals IndexVals AI53.0†70.5
Vals Finance IndexVals AI52.9%†67.0%
Vals Router IndexVals AI53.0%†–
Metadata
AA Openness IndexArtificial Analysis44.4†–

Data as of 2026-11-09. Best configuration of each model; every score links to its source on the model pages.

DuelLab GameBench 2, release gb2-20261103-eaea31df94e6 (https://benchmarks.duellab.org/releases/gb2-20261103-eaea31df94e6/), CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Scores mapped to BenchLeader model names; not endorsed by DuelLab.

What changed

  • DeepSeek V4 Pro (high): BenchLeader Index 57.5 → 59.1.
  • Gemini 4 Carbon (high): BenchLeader Index 69.3 → 68.5.
  • Gemini 4 Carbon (high): BenchLeader Index 68.8 → 69.3.
  • Gemini 4 Carbon (high): 1 score revision.
  • DeepSeek V4 Pro (high): 1 score revision.
  • DeepSeek V4 Pro (high): 1 score revision.
  • Gemini 4 Carbon (high): 2 score revisions.
  • Gemini 4 Carbon (high): 1 score revision.
  • Gemini 4 Carbon (high): 2 score revisions.
  • Gemini 4 Carbon (high): 1 score revision.

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

DeepSeek V4 Pro vs Gemini 4 Carbon: questions

Is Gemini 4 Carbon better than DeepSeek V4 Pro?
Gemini 4 Carbon (high) leads on quality: 68.5 vs 59.1. The BenchLeader Index combines every independent quality benchmark; Gemini 4 Carbon (high) is ahead overall as of 9 Nov 2026, but check the category scores for your use.
DeepSeek V4 Pro vs Gemini 4 Carbon: which is cheaper, and will that change?
DeepSeek V4 Pro is cheaper today: $0.990 against $8.00 per million tokens, blended at three input tokens per output token. No price change has been announced for either.
Is DeepSeek V4 Pro or Gemini 4 Carbon better for coding?
Gemini 4 Carbon scores higher in coding (75 vs 54 on the category index, where 50 is average). On Arena WebDev, the top shared coding benchmark, it is 1581 for DeepSeek V4 Pro against 1768 for Gemini 4 Carbon.
Which is faster, DeepSeek V4 Pro or Gemini 4 Carbon?
Gemini 4 Carbon streams ~1.5× faster (75 vs 44 output tokens per second, Artificial Analysis 7-day median).
Which is better for long documents?
Both take about 1M tokens: DeepSeek V4 Pro 1,048,576, Gemini 4 Carbon 1,000,000.
More questions (3)
Which has lower latency?
DeepSeek V4 Pro answers first: 1.5 s against 69.4 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 DeepSeek V4 Pro vs Gemini 4 Carbon?
Across the 8 benchmarks both have, DeepSeek V4 Pro leads 0, Gemini 4 Carbon leads 8. Every overall index that rates both puts Gemini 4 Carbon ahead.
When was Gemini 4 Carbon released?
Gemini 4 Carbon was released on 28 Oct 2026.