BenchLeader

DeepSeek V4 Flash vs DeepSeek V4 Pro

Full profiles: DeepSeek V4 Flash (DeepSeek) and DeepSeek V4 Pro (DeepSeek): every benchmark, price and rank.

Updated · Data from 8 sources

As of 16 Nov 2026, DeepSeek V4 Flash and DeepSeek V4 Pro are neck and neck on quality (57.2 vs 56.0 on the BenchLeader Index, inside the ±5.9 margin), and DeepSeek V4 Flash is 3.8× cheaper at $0.262 vs $0.990 per 1M tokens blended. Choose V4 Flash for knowledge (55 vs 41), reasoning (59 vs 56), and coding (55 vs 52) and fast first answers (0.7 s vs 42.8 s), where it also streams faster; choose V4 Pro for human preference (65 vs 62) and maths (65 vs 62).

Key facts (as of 16 Nov 2026)

  • DeepSeek V4 Flash scores 57.2 on the BenchLeader Index and DeepSeek V4 Pro scores 56.0 (BenchLeader, 16 Nov 2026).
  • DeepSeek V4 Flash costs $0.262 per 1M tokens blended (list price, 16 Nov 2026).
  • DeepSeek V4 Pro costs $0.990 per 1M tokens blended (list price, 16 Nov 2026).
  • DeepSeek V4 Flash streams faster than DeepSeek V4 Pro (Artificial Analysis, 7-day median, 16 Nov 2026).
  • Of the 15 benchmarks both models have, DeepSeek V4 Pro leads 13 (BenchLeader, 16 Nov 2026).
  • On Arena WebDev, DeepSeek V4 Flash is rated 1582 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
CodingDeepSeek V4 Flash55 vs 52 category score; Arena WebDev 1582 vs 1581Medium
MathsDeepSeek V4 Pro65 vs 62 category score; Arena Maths 1465 vs 1441Medium
ReasoningDeepSeek V4 Flash59 vs 56 category score; ARC-AGI-2 56.0% vs 59.7%Medium
KnowledgeDeepSeek V4 Flash55 vs 41 category scoreHigh
Chat quality (human preference)DeepSeek V4 Pro65 vs 62 on Arena votesMedium
Lowest costDeepSeek V4 Flash$0.262 vs $1.98 per 1M blended; agentic coding session $0.011 vs $0.095–
Fastest first answerDeepSeek V4 Flash0.7 s (DeepSeek V4 Flash (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

  • DeepSeek V4 Flash (high) and DeepSeek V4 Pro (high) are neck and neck on quality (57.2 vs 56.0, within the ±5.9 margin).
  • DeepSeek V4 Flash (high) is stronger in coding, knowledge, reasoning.
  • DeepSeek V4 Pro (high) is stronger in human preference, maths.
  • DeepSeek V4 Flash (high) is 3.8× cheaper ($0.262 vs $0.990 per 1M blended).

At a glance

SpecDeepSeek V4 FlashDeepSeek V4 Pro
DeveloperDeepSeekDeepSeek
Released31 Jul 202613 Aug 2026
Context window (tokens)1,048,5761,048,576
Compared atDeepSeek V4 Flash (high)DeepSeek V4 Pro (high)
Open weightsYesYes
Price in / out / cached input per 1M$0.150 / $0.600 / $0.0030$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.

CategoryDeepSeek V4 Flash (high)DeepSeek V4 Pro (high)Leader
Reasoning59.255.6DeepSeek V4 Flash
Coding55.051.5DeepSeek V4 Flash
Maths61.964.5DeepSeek V4 Pro
Knowledge54.841.2DeepSeek V4 Flash
Human preference62.065.0DeepSeek V4 Pro
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 avg75ReasoningCodingMathsKnowledgeHumanpreferenceDeepSeek V4 Flash, Reasoning: 59.2DeepSeek V4 Flash, Coding: 55.0DeepSeek V4 Flash, Maths: 61.9DeepSeek V4 Flash, Knowledge: 54.8DeepSeek V4 Flash, Human preference: 62.0DeepSeek V4 Pro, Reasoning: 55.6DeepSeek V4 Pro, Coding: 51.5DeepSeek V4 Pro, Maths: 64.5DeepSeek V4 Pro, Knowledge: 41.2DeepSeek V4 Pro, Human preference: 65.0

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

Model, click to add or removeIndex

Where sources disagree

Across the 15 benchmarks both have, DeepSeek V4 Flash leads 2, DeepSeek V4 Pro leads 13. Every overall index that rates both puts DeepSeek V4 Pro ahead.

Overall indexDeepSeek V4 Flash (high)DeepSeek V4 Pro (high)Leader
BenchLeader Index57.256.0neck and neck
Arena Text14391464DeepSeek 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.

MeasureDeepSeek V4 Flash (high)DeepSeek V4 Pro (high)
Blended price per 1M tokens$0.262$1.98
Input / output per 1M$0.150 / $0.600$1.32 / $3.96
Cached input per 1M$0.0030$0.022
Price per task: chat reply$0.0002$0.0017
Price per task: summarize a 30-page report$0.0022$0.018
Price per task: code edit$0.0018$0.014
Price per task: agentic coding session$0.011$0.095
Price per task: structured extraction$0.0004$0.0034
Agentic coding session, input mostly cached$0.0048$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 speed82 tok/s80 tok/s
Time to first answer0.7 s42.8 s

Each model is compared at its best-ranked reasoning-effort setting. DeepSeek V4 Flash (high) scores 57.2 and takes 1 s to its first answer token; DeepSeek V4 Flash (max) scores 56.8 in 28 s, DeepSeek V4 Flash scores 53.8 in 1 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

125 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 Flash (high)DeepSeek V4 Pro (high)
BenchLeader IndexBenchLeader57.256.0
Coding scoreBenchLeader55.051.5
Human preference scoreBenchLeader62.065.0
Knowledge scoreBenchLeader54.841.2
Maths scoreBenchLeader61.964.5
Reasoning scoreBenchLeader59.255.6
Long context scoreBenchLeader–64.3
Blended price $/MBenchLeader$0.262$1.98
Output speedBenchLeader82 tok/s80 tok/s
Time to first answerBenchLeader0.7 s42.8 s
Context windowBenchLeader1,048,576 tokens1,048,576 tokens
Reasoning
GPQA DiamondEpoch AI Benchmarking Hub91.0%†90.9%
SimpleBenchSimpleBench61.1%†–
Arena Hard PromptsArena (LMArena)14591482
LiveBench 1.0 ReasoningLiveBench86.6%†85.8%†
GPQA Diamond (AA)Artificial Analysis90.8%†92.8%†
Humanity's Last Exam (AA)Artificial Analysis38.5%†41.5%
HLE-Diamond (AA)Artificial Analysis–20.5%
GPQA Diamond (Vals)Vals AI89.9%92.4%†
Kagi LLM BenchmarkKagi LLM Benchmark52.2%†53.5%†
ARC-AGI-1ARC Prize87.0%87.2%
ARC-AGI-2ARC Prize56.0%59.7%
CritPtArtificial Analysis16.6%†18.6%
CritPt v2 hard subset (AA)Artificial Analysis–28.5%
HLE-Diamond (Scale)Scale AI SEAL–13.4%†
Arena ExpertArena (LMArena)14671481
Chess PuzzlesEpoch AI Benchmarking Hub32.0%†16.0%
Mystery Game PuzzlesEpoch AI Benchmarking Hub34.0%†43.0%†
CL-bench LifeEpoch AI Benchmarking Hub–13.5%
LMCAEpoch AI Benchmarking Hub41.7%†45.5%†
DTBenchEpoch AI Benchmarking Hub90.9%†93.9%†
ForecastBenchForecastBench–60.7%†
Coding
SWE-bench Verified (Epoch)Epoch AI Benchmarking Hub–77.6%†
SciCodeSciCode42.0%46.4%
WeirdMLWeirdML57.0%46.5%
FrontierCodeCognition18.8%†28.6%
Show all 125 benchmarks
MetricPublisherDeepSeek V4 Flash (high)DeepSeek V4 Pro (high)
Coding
Arena CodingArena (LMArena)14801503
Arena WebDevArena (LMArena)15821581
LiveBench 1.0 CodingLiveBench75.0%†77.2%†
SciCode (AA)Artificial Analysis50.4%†51.7%
LiveCodeBenchVals AI87.3%87.5%†
IOIVals AI32.7%51.6%†
SWE-bench (Vals)Vals AI88.8%96.4%†
Code MigrationVals AI38.6%41.5%†
Code Migration v1.1Vals AI–52.0%†
ProgramBenchVals AI0.0%0.0%†
Vibe Code Bench 1-100Vals AI–17.5%†
Vibe Code Bench v1.1Vals AI74.7%82.3%†
SWE Atlas v1.0: Codebase QnAScale AI SEAL–27.1%†
SWE Atlas v1.0: Test WritingScale AI SEAL–27.1%†
MirrorCodeEpoch AI Benchmarking Hub–33.3%†
Surface Evolver BenchEpoch AI Benchmarking Hub–40.0%
ALE-BenchEpoch AI Benchmarking Hub678.21006.1
Terminal-Bench 4.0 (AA)Artificial Analysis12.1%†6.1%
Terminal-Bench 2.1 (AA)Artificial Analysis78.7%†78.7%†
Startrise FrontendStartrise–62.7%†
GameBench 2 OverallDuelLab GameBench 234.1†41.9†
GameBench 2 by reasoning settingDuelLab GameBench 243.445.1
Design Arena WebsiteDesign Arena1199†1227†
FrontierCode 1.1 (AA)Artificial Analysis–38.8%†
FrontierCode 2.0 (AA)Artificial Analysis–16.7%
Agents & tools
Terminal-BenchTerminal-Bench–6.7%†
APEX-AgentsMercor–47.3%†
LiveBench Agentic CodingLiveBench46.8%†55.0%†
Terminal-Bench 2.1 (Vals)Vals AI67.0%†50.2%†
HiL-BenchScale AI SEAL47.8%†38.5%†
GDPval-AA v2.1, normalised scoreArtificial Analysis–47.5%†
GDPval-AA v2.1Artificial Analysis1437†1336
τ³-Banking (AA)Artificial Analysis39.4%†39.6%†
CyberBenchVals AI72.1%72.9%†
Finance Agent v2Vals AI49.5%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 AI8.3%7.5%†
Legal Research BenchVals AI30.3%40.9%†
SkillsBenchVals AI50.7%53.8%†
SREBenchVals AI–1.9%†
Tax Agent BenchVals AI28.1%26.8%†
Terminal-Bench 2.0 (Vals)Vals AI–56.2%†
Terminal-Bench 4.0 (Vals)Vals AI18.7%14.1%†
Terminal-Bench ScienceVals AI4.3%4.3%†
Vending-Bench 2Andon Labs–3284.5†
AutomationBenchArtificial Analysis54.0%†46.2%
GDP.pdf (AA)Artificial Analysis11.0%†10.4%
EnterpriseOps-GymArtificial Analysis–49.6%†
AA-Briefcase v1.1Artificial Analysis1255†1148
Design Arena AgenticDesign Arena–927†
Terminal-Bench-Science 0.1 (AA)Artificial Analysis–1.9%
Maths
FrontierMath Tiers 1–3Epoch AI Benchmarking Hub57.5%†64.6%†
FrontierMath Tier 4Epoch AI Benchmarking Hub24.4%†26.8%†
OTIS Mock AIMEEpoch AI Benchmarking Hub94.4%†95.6%
ProofBenchVals AI56.0%†50.0%†
LiveBench MathematicsLiveBench86.8%†95.1%†
AIME 2026MathArena95.8%†96.7%†
HMMT February 2026MathArena93.9%†93.9%†
USAMO 2026MathArena–60.7%†
MathArena ApexMathArena27.1%†28.1%†
Arena MathsArena (LMArena)14411465
Knowledge
SimpleQA VerifiedEpoch AI Benchmarking Hub33.6%†52.9%†
LiveBench 1.0 Data AnalysisLiveBench79.3%†79.2%†
AA-Omniscience v1.1Artificial Analysis–-6.5
MMLU-ProVals AI86.2%87.3%†
LegalBenchVals AI77.7%82.4%†
CorpFinVals AI61.9%65.4%†
TaxEvalVals AI70.7%73.1%†
BioMysteryBenchVals AI64.4%–
CaseLaw v2Vals AI–59.4%†
Excel Modeling BenchmarkVals AI57.0%52.8%†
Excel Modeling Benchmark v1.1Vals AI–65.1%†
MedCodeVals AI41.4%42.5%†
MedScribeVals AI80.4%80.2%†
Public Benefits Bench v1.1Vals AI–62.9%†
Immigration BenchVals AI–61.7%†
DrugDiscoveryBenchScale AI SEAL–31.7%†
AA-Omniscience v1.1: accuracyArtificial Analysis–45.1%
AA-Omniscience v1.1: non-hallucinationArtificial Analysis–5.9%
Instruction following
LiveBench LanguageLiveBench79.2%†82.1%†
LiveBench Instruction FollowingLiveBench65.5%†67.7%†
Arena Instruction FollowingArena (LMArena)14341461
Human preference
Arena TextArena (LMArena)14391464
EQ-Bench 4EQ-Bench–1156†
EQ-Bench Creative Writing v3EQ-Bench1559†1553†
Arena Creative WritingArena (LMArena)14111444
Arena Multi-turnArena (LMArena)14471469
Arena Longer QueriesArena (LMArena)14491474
Long context
AA-LCR v1.1Artificial Analysis79.7%†79.3%
Composite
Epoch Capabilities IndexEpoch AI Benchmarking Hub154.4†155.9†
LiveBench 1.0LiveBench74.2%†77.4%†
AA Intelligence Index v5.0.1Artificial Analysis–20.8
AA Intelligence Index v4.3.2Artificial Analysis34.3†36†
AA Progress Index v1Artificial Analysis309.4†311.5†
Vals IndexVals AI–53.0†
Vals Finance IndexVals AI–52.9%†
Vals Router IndexVals AI–53.0%†
Vals Legal IndexVals AI–69.0%†
Metadata
AA Openness IndexArtificial Analysis44.4†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

  • DeepSeek V4 Flash (high): 1 score revision.
  • 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).
  • DeepSeek V4 Pro (high): BenchLeader Index 57.5 → 59.1.
  • DeepSeek V4 Pro (high): 1 score revision.
  • DeepSeek V4 Flash (high): 1 score revision.
  • DeepSeek V4 Pro (high): 1 score revision.
  • DeepSeek V4 Flash (high): 1 new result (GameBench 2 by reasoning setting).
  • DeepSeek V4 Flash (high): 1 new result (Harvey's Legal Agent Benchmark v1).
  • DeepSeek V4 Flash (high): BenchLeader Index 59.3 → 57.3.

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

DeepSeek V4 Flash vs DeepSeek V4 Pro: questions

Is DeepSeek V4 Flash better than DeepSeek V4 Pro?
DeepSeek V4 Flash (high) and DeepSeek V4 Pro (high) are neck and neck on quality (57.2 vs 56.0, within the ±5.9 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.
DeepSeek V4 Flash vs DeepSeek V4 Pro: which is cheaper, and will that change?
DeepSeek V4 Flash is cheaper today: $0.262 against $0.990 per million tokens, blended at three input tokens per output token. No price change has been announced for either.
Is DeepSeek V4 Flash or DeepSeek V4 Pro better for coding?
DeepSeek V4 Flash scores higher in coding (55 vs 52 on the category index, where 50 is average). On Arena WebDev, the top shared coding benchmark, it is 1582 for DeepSeek V4 Flash against 1581 for DeepSeek V4 Pro.
Which is faster, DeepSeek V4 Flash or DeepSeek V4 Pro?
DeepSeek V4 Flash streams faster (82 vs 80 output tokens per second, Artificial Analysis 7-day median).
Which is better for long documents?
Both accept 1,048,576 tokens of context.
More questions (3)
Which has lower latency?
DeepSeek V4 Flash answers first: 0.7 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 DeepSeek V4 Flash vs DeepSeek V4 Pro?
Across the 15 benchmarks both have, DeepSeek V4 Flash leads 2, DeepSeek V4 Pro leads 13. 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.