BenchLeader

DeepSeek V4 Pro vs Gemini 3.8 Flash

Full profiles: DeepSeek V4 Pro (DeepSeek) and Gemini 3.8 Flash (Google): every benchmark, price and rank.

Updated · Data from 8 sources

As of 16 Nov 2026, Gemini 3.8 Flash is ahead of DeepSeek V4 Pro on quality (60.2 vs 56.0 on the BenchLeader Index), but DeepSeek V4 Pro is 1.5× cheaper at $0.990 vs $1.50 per 1M tokens blended. Gemini 3.8 Flash leads in every category we measure both in. Gemini 3.8 Flash's current price runs until 31 Dec 2026, then $3.00.

Key facts (as of 16 Nov 2026)

  • DeepSeek V4 Pro scores 56.0 on the BenchLeader Index and Gemini 3.8 Flash scores 60.2 (BenchLeader, 16 Nov 2026).
  • DeepSeek V4 Pro costs $0.990 per 1M tokens blended (list price, 16 Nov 2026).
  • Gemini 3.8 Flash costs $1.50 per 1M tokens blended until 31 Dec 2026, then $3.00 (https://ai.google.dev/gemini-api/docs/pricing, checked 20 Oct 2026).
  • Gemini 3.8 Flash streams ~2× faster than DeepSeek V4 Pro (Artificial Analysis, 7-day median, 16 Nov 2026).
  • Of the 21 benchmarks both models have, Gemini 3.8 Flash leads 19 (BenchLeader, 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
CodingEither52 vs 52 category score; FrontierCode 2.0 (AA) 16.7% vs 16.5%Either
ReasoningGemini 3.8 Flash59 vs 56 category score; HLE-Diamond (AA) 29.1% vs 20.5%Medium
KnowledgeGemini 3.8 Flash58 vs 41 category score; AA-Omniscience v1.1 23.7 vs -6.5High
Long documentsGemini 3.8 Flash67 vs 64 category score; AA-LCR v1.1 84.0% vs 79.3%Medium
Chat quality (human preference)–Only DeepSeek V4 Pro has an Arena rating (65 category score)Not comparable
Lowest costGemini 3.8 Flash (until 31 Dec 2026)$1.50 vs $1.98 per 1M blended; agentic coding session $0.060 vs $0.095; from 1 Jan 2027: $3.00 vs $1.98–
Fastest first answerGemini 3.8 Flash14.8 s (Gemini 3.8 Flash (medium)) 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

  • Gemini 3.8 Flash (medium) leads on quality: 60.2 vs 56.0.
  • Gemini 3.8 Flash (medium) is stronger in coding, knowledge, long context, reasoning.
  • DeepSeek V4 Pro (high) is 1.5× cheaper ($0.990 vs $1.50 per 1M blended) until 31 Dec 2026, then $1.98 vs $3.00 from 1 Jan 2027.
  • Gemini 3.8 Flash (medium) streams 1.4× faster (116 vs 80 tokens per second).

At a glance

SpecDeepSeek V4 ProGemini 3.8 Flash
DeveloperDeepSeekGoogle
Released13 Aug 20262 Sep 2026
Context window (tokens)1,048,5761,048,576
Compared atDeepSeek V4 Pro (high)Gemini 3.8 Flash (medium)
Open weightsYesNo
Price in / out / cached input per 1M$1.32 / $3.96 / $0.022$0.750 / $3.75 / $0.075
Announced price changeNone announced$1.50 / $7.50 from 1 Jan 2027

Limits, availability and announced prices from the vendors' own pages: ai.google.dev pricing (checked 20 Oct 2026). Everything else is from the day's data.

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 3.8 Flash (medium)Leader
Reasoning55.658.7Gemini 3.8 Flash
Coding51.551.6level
Agents & tools–75.0not comparable
Maths64.5–not comparable
Knowledge41.257.7Gemini 3.8 Flash
Human preference65.0–not comparable
Multimodal–66.7not comparable
Long context64.366.7Gemini 3.8 Flash

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 avg75ReasoningCodingKnowledgeLong contextDeepSeek V4 Pro, Reasoning: 55.6DeepSeek V4 Pro, Coding: 51.5DeepSeek V4 Pro, Knowledge: 41.2DeepSeek V4 Pro, Long context: 64.3Gemini 3.8 Flash, Reasoning: 58.7Gemini 3.8 Flash, Coding: 51.6Gemini 3.8 Flash, Knowledge: 57.7Gemini 3.8 Flash, Long context: 66.7

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

Model, click to add or removeIndex

Where sources disagree

Across the 21 benchmarks both have, DeepSeek V4 Pro leads 2, Gemini 3.8 Flash leads 19. Every overall index that rates both puts Gemini 3.8 Flash ahead.

Overall indexDeepSeek V4 Pro (high)Gemini 3.8 Flash (medium)Leader
BenchLeader Index56.060.2Gemini 3.8 Flash
AA Intelligence Index v5.0.120.830.1Gemini 3.8 Flash
Vals Index53.0†56.0Gemini 3.8 Flash
Arena Text14641497†Gemini 3.8 Flash

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 3.8 Flash (medium)
Blended price per 1M tokens, until 31 Dec 2026$1.98$1.50
Input / output per 1M$1.32 / $3.96$0.750 / $3.75
Blended price per 1M from 1 Jan 2027$1.98$3.00
Input / output per 1M from 1 Jan 2027$1.32 / $3.96$1.50 / $7.50
Cached input per 1M$0.022$0.075
Price per task: chat reply$0.0017$0.0014
Price per task: summarize a 30-page report$0.018$0.011
Price per task: code edit$0.014$0.010
Price per task: agentic coding session$0.095$0.060
Price per task: structured extraction$0.0034$0.0023
Agentic coding session, input mostly cached$0.037$0.030
Cost per benchmark task (AA Index)$0.849$6.26
Output tokens per task (AA Index)70k82k
Of which reasoning45k47k
Time per task (AA Index)15 min12 min
Output speed80 tok/s116 tok/s
Time to first answer42.8 s14.8 s
Time to a complete answer, last measured 13 Nov 2026–41.8 s

Per token is not per task: an average task of Artificial Analysis' Intelligence Index cost $0.849 on DeepSeek V4 Pro (high) and $6.26 on Gemini 3.8 Flash (medium), and Gemini 3.8 Flash (medium) wrote 1.2× as many output tokens for the same work.

Each model is compared at its best-ranked reasoning-effort setting. 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. Gemini 3.8 Flash (medium) scores 60.2 and takes 15 s to its first answer token; Gemini 3.8 Flash (high) scores 59.4 in 20 s, Gemini 3.8 Flash scores 56.4 in 2 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

191 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 3.8 Flash (medium)
BenchLeader IndexBenchLeader56.060.2
Coding scoreBenchLeader51.551.6
Human preference scoreBenchLeader65.0–
Knowledge scoreBenchLeader41.257.7
Long context scoreBenchLeader64.366.7
Maths scoreBenchLeader64.5–
Reasoning scoreBenchLeader55.658.7
Agents & tools scoreBenchLeader–75.0
Multimodal scoreBenchLeader–66.7
Blended price $/MBenchLeader$1.98$1.50
Output speedBenchLeader80 tok/s116 tok/s
Time to first answerBenchLeader42.8 s14.8 s
Context windowBenchLeader1,048,576 tokens1,048,576 tokens
Reasoning
GPQA DiamondEpoch AI Benchmarking Hub90.9%95.4%†
Humanity's Last ExamScale AI / CAIS–44.5%†
SimpleBenchSimpleBench–82.4%†
Arena Hard PromptsArena (LMArena)14821518†
LiveBench 2.0 ReasoningLiveBench–89.3%†
LiveBench 1.0 ReasoningLiveBench85.8%†89.3%†
GPQA Diamond (AA)Artificial Analysis92.8%†93.5%
Humanity's Last Exam (AA)Artificial Analysis41.5%42.1%
HLE-Diamond (AA)Artificial Analysis20.5%29.1%
GPQA Diamond (Vals)Vals AI92.4%†94.4%†
Kagi LLM BenchmarkKagi LLM Benchmark53.5%†–
ARC-AGI-1ARC Prize87.2%97.5%
ARC-AGI-2ARC Prize59.7%82.9%
ARC-AGI-3ARC Prize–24.2%
CritPtArtificial Analysis18.6%12.3%
CritPt v2 hard subset (AA)Artificial Analysis28.5%45.0%
MysteryMechanismVals AI–36.5%†
HLE-Diamond (Scale)Scale AI SEAL13.4%†34.3%†
Arena ExpertArena (LMArena)14811527†
Chess PuzzlesEpoch AI Benchmarking Hub16.0%58.0%†
Mystery Game PuzzlesEpoch AI Benchmarking Hub43.0%†47.0%†
CL-bench LifeEpoch AI Benchmarking Hub13.5%–
LMCAEpoch AI Benchmarking Hub45.5%†52.9%
DTBenchEpoch AI Benchmarking Hub93.9%†94.9%
ForecastBenchForecastBench60.7%†–
Show all 191 benchmarks
MetricPublisherDeepSeek V4 Pro (high)Gemini 3.8 Flash (medium)
Coding
SWE-bench Verified (Epoch)Epoch AI Benchmarking Hub77.6%†–
SciCodeSciCode46.4%55.1%
WeirdMLWeirdML46.5%84.8%†
FrontierCodeCognition28.6%41.2%
Arena CodingArena (LMArena)15031531†
Arena WebDevArena (LMArena)15811584†
LiveBench 2.0 CodingLiveBench–73.3%†
LiveBench 1.0 CodingLiveBench77.2%†72.5%†
SciCode (AA)Artificial Analysis51.7%55.1%
LiveCodeBenchVals AI87.5%†89.5%†
IOIVals AI51.6%†56.9%†
SWE-bench (Vals)Vals AI96.4%†80.0%†
Code MigrationVals AI41.5%†36.5%†
Code Migration v1.1Vals AI52.0%†38.6%†
ProgramBenchVals AI0.0%†1.0%†
SWE-bench MultimodalVals AI–40.8%†
Vibe Code Bench 1-100Vals AI17.5%†18.8%†
Vibe Code Bench v1.1Vals AI82.3%†78.7%†
SWE Atlas v1.0: Codebase QnAScale AI SEAL27.1%†47.0%†
SWE Atlas v1.0: RefactoringScale AI SEAL–44.8%†
SWE Atlas v1.0: Test WritingScale AI SEAL27.1%†53.7%†
SWE Atlas v1.1Scale AI SEAL–57.3%†
SWE Atlas v1.1: Codebase QnAScale AI SEAL–45.2%†
SWE Atlas v1.1: Test WritingScale AI SEAL–71.1%†
SWE Atlas v1.1: RefactoringScale AI SEAL–55.7%†
SWE-Bench Pro V2Scale AI SEAL–94.9%†
SWE-Bench Pro V2 HARDScale AI SEAL–58.8%†
MirrorCodeEpoch AI Benchmarking Hub33.3%†38.9%†
Surface Evolver BenchEpoch AI Benchmarking Hub40.0%76.9%†
DeepSWE v1.1Epoch AI Benchmarking Hub–71.0%
CursorBenchCursor–37.3%
ALE-BenchEpoch AI Benchmarking Hub1006.11269.7†
FrontierSWEEpoch AI Benchmarking Hub–19.6%†
WeirdML v3WeirdML–7.4%†
Terminal-Bench 4.0 (AA)Artificial Analysis6.1%19.7%
Terminal-Bench 2.1 (AA)Artificial Analysis78.7%†83.9%
Startrise FrontendStartrise62.7%†64.6%†
GameBench 2 OverallDuelLab GameBench 241.9†59.0†
GameBench 2 by reasoning settingDuelLab GameBench 245.159.7
Design Arena WebsiteDesign Arena1227†1291†
FrontierCode 1.1 (AA)Artificial Analysis38.8%†37.1%
FrontierCode 2.0 (AA)Artificial Analysis16.7%16.5%
Agents & tools
Terminal-BenchTerminal-Bench6.7%†18.2%†
Remote Labor IndexScale AI / CAIS–5.8%†
APEX-AgentsMercor47.3%†64.3%†
Arena AgentArena (LMArena)–2.3†
LiveBench Long AgenticLiveBench–16.5%†
LiveBench BrowserLiveBench–45.1%†
LiveBench Agentic CodingLiveBench55.0%†54.2%†
Terminal-Bench 2.1 (Vals)Vals AI50.2%†81.3%†
HiL-BenchScale AI SEAL38.5%†41.5%†
GDPval-AA v2.1, normalised scoreArtificial Analysis47.5%†46.1%
GDPval-AA v2.1Artificial Analysis13361432
τ³-Banking (AA)Artificial Analysis39.6%†45.8%
ITBench SRE (AA)Artificial Analysis–52.5%†
Analyst Agent (AA)Artificial Analysis–46.3%†
CUA-benchVals AI–4.2%†
CyberBenchVals AI72.9%†43.8%†
Finance Agent v2Vals AI50.4%†61.4%†
Finance Agent v2.1Vals AI57.1%†65.8%†
Harvey's Legal Agent Benchmark v1.1Vals AI10.8%†12.1%†
Harvey's Legal Agent Benchmark v1Vals AI7.5%†10.0%†
Legal Research BenchVals AI40.9%†38.9%†
SkillsBenchVals AI53.8%†58.0%†
SREBenchVals AI1.9%†11.1%†
Tax Agent BenchVals AI26.8%†32.1%†
Terminal-Bench 2.0 (Vals)Vals AI56.2%†–
Terminal-Bench 4.0 (Vals)Vals AI14.1%†19.7%†
Terminal-Bench ScienceVals AI4.3%†8.6%†
Time Horizon Index: KSPVals AI–18.8%†
Arena Agent Task CompletionArena (LMArena)–1.3†
Vending-Bench 2Andon Labs3284.5†5093.8†
AutomationBenchArtificial Analysis46.2%60.9%
GDP.pdf (AA)Artificial Analysis10.4%22.8%
Harvey LAB-AA v1.1Harvey–0.0%†
EnterpriseOps-GymArtificial Analysis49.6%†–
AA-Briefcase v1.1Artificial Analysis11481183
Design Arena AgenticDesign Arena927†1225†
Terminal-Bench-Science 0.1 (AA)Artificial Analysis1.9%10.5%
CWE-Bench (AA)Artificial Analysis–52.5%†
DeepSec-Bench (AA)Artificial Analysis–14.5%†
CyberGym end-to-end (AA)Artificial Analysis–8.4%†
MKA (AA)Artificial Analysis–35.3%†
FACTS SearchGoogle–80.4%†
Maths
FrontierMath Tiers 1–3Epoch AI Benchmarking Hub64.6%†68.4%†
FrontierMath Tier 4Epoch AI Benchmarking Hub26.8%†21.9%†
OTIS Mock AIMEEpoch AI Benchmarking Hub95.6%98.9%†
ProofBenchVals AI50.0%†48.0%†
LiveBench MathematicsLiveBench95.1%†91.6%†
AIME 2026MathArena96.7%†–
HMMT February 2026MathArena93.9%†–
USAMO 2026MathArena60.7%†–
MathArena ApexMathArena28.1%†–
Arena MathsArena (LMArena)14651519†
FrontierMath ErdősEpoch AI Benchmarking Hub–0.0%†
Knowledge
SimpleQA VerifiedEpoch AI Benchmarking Hub52.9%†69.7%†
LiveBench 2.0 Data AnalysisLiveBench–54.0%†
LiveBench ResearchLiveBench–79.4%†
LiveBench 1.0 Data AnalysisLiveBench79.2%†54.0%†
AA-Omniscience v1.1Artificial Analysis-6.523.7
MMLU-ProVals AI87.3%†90.2%†
LegalBenchVals AI82.4%†87.0%†
CorpFinVals AI65.4%†–
TaxEvalVals AI73.1%†74.5%†
PRBench FinanceScale AI SEAL–49.5%†
PRBench LegalScale AI SEAL–48.9%†
BioMysteryBenchVals AI–62.2%†
CaseLaw v2Vals AI59.4%†–
Excel Modeling BenchmarkVals AI52.8%†72.2%†
Excel Modeling Benchmark v1.1Vals AI65.1%†72.9%†
MedCodeVals AI42.5%†48.1%†
MedScribeVals AI80.2%†84.5%†
Public Benefits Bench v1.1Vals AI62.9%†65.3%†
Access to Justice BenchmarkVals AI–67.1%†
Financial Coaching BenchVals AI–66.3%†
Immigration BenchVals AI61.7%†63.4%†
DrugDiscoveryBenchScale AI SEAL31.7%†–
GDP.pdfEpoch AI Benchmarking Hub–23.4%
AA-Omniscience v1.1: accuracyArtificial Analysis45.1%49.6%
AA-Omniscience v1.1: non-hallucinationArtificial Analysis5.9%48.5%
HealthBench Professional (AA)OpenAI–51.7%†
FACTS ParametricGoogle–78.7%†
Instruction following
LiveBench LanguageLiveBench82.1%†87.8%†
LiveBench Instruction FollowingLiveBench67.7%†81.4%†
Arena Instruction FollowingArena (LMArena)14611492†
Human preference
Arena TextArena (LMArena)14641497†
EQ-Bench 4EQ-Bench1156†–
EQ-Bench Creative Writing v3EQ-Bench1553†1744†
Arena Creative WritingArena (LMArena)14441488†
Arena Multi-turnArena (LMArena)14691505†
Arena Longer QueriesArena (LMArena)14741511†
Multimodal
Arena VisionArena (LMArena)–1317†
LiveBench DesignLiveBench–69.7%†
MMMU-ProArtificial Analysis–84.2%
MMMU-Pro (Vals)Vals AI–89.1%†
MortgageTaxVals AI–65.3%†
SAGEVals AI–88.2%†
AudioMCScale AI SEAL–60.4%†
AudioMC (text output)Scale AI SEAL–60.4%†
Humanity's Sixth SenseScale AI SEAL–41.6%†
GeoBenchGeoBench–86.0%†
Blueprint-Bench 2Epoch AI Benchmarking Hub–38.6%†
Furniture AssemblyEpoch AI Benchmarking Hub–31.7%†
VIALS-AAHandshake AI–41.0%†
Code2Video-AAArtificial Analysis–899†
AA-3DArtificial Analysis–974†
FACTS MultimodalGoogle–57.6%†
Long context
AA-LCR v1.1Artificial Analysis79.3%84.0%
Arena DocumentArena (LMArena)–1469†
MLCRArtificial Analysis–12.8%
FACTS GroundingGoogle–68.2%†
Composite
Epoch Capabilities IndexEpoch AI Benchmarking Hub155.9†158.8†
LiveBench 2.0LiveBench–57.3%†
LiveBench 1.0LiveBench77.4%†75.8%†
AA Intelligence Index v5.0.1Artificial Analysis20.830.1
AA Intelligence Index v4.3.2Artificial Analysis36†39.8
AA Progress Index v1Artificial Analysis311.5†319.1
Vals IndexVals AI53.0†56.0†
Vals Finance IndexVals AI52.9%†60.5%†
Vals RSI IndexVals AI–19.9%†
Vals Science IndexVals AI–35.8%†
Vals Router IndexVals AI53.0%†56.0%†
Vals Legal IndexVals AI69.0%†69.8%†
AA Cyber IndexArtificial Analysis–25.1†
Safety & honesty
DistressBenchScale AI SEAL–70.7%†
Metadata
AA Openness IndexArtificial Analysis44.4–

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

  • Gemini 3.8 Flash: list price moves to $1.50 / $7.50 per 1M input / output tokens (announced).
  • Gemini 3.8 Flash (medium): 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 Pro (high): 1 score revision.
  • Gemini 3.8 Flash (medium): 2 score revisions.
  • Gemini 3.8 Flash (medium): 1 score revision.
  • Gemini 3.8 Flash (medium): 1 new result (AA-Briefcase v1.1).

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

DeepSeek V4 Pro vs Gemini 3.8 Flash: questions

Is Gemini 3.8 Flash better than DeepSeek V4 Pro?
Gemini 3.8 Flash (medium) leads on quality: 60.2 vs 56.0. The BenchLeader Index combines every independent quality benchmark; Gemini 3.8 Flash (medium) is ahead overall as of 16 Nov 2026, but check the category scores for your use.
DeepSeek V4 Pro vs Gemini 3.8 Flash: which is cheaper, and will that change?
DeepSeek V4 Pro is cheaper today: $0.990 against $1.50 per million tokens, blended at three input tokens per output token. Gemini 3.8 Flash's current price runs until 31 Dec 2026, then $3.00.
Is DeepSeek V4 Pro or Gemini 3.8 Flash better for coding?
They are level in coding (52 vs 52 on the category index). On FrontierCode 2.0 (AA), the top shared coding benchmark, it is 16.7% for DeepSeek V4 Pro against 16.5% for Gemini 3.8 Flash.
Which is faster, DeepSeek V4 Pro or Gemini 3.8 Flash?
Gemini 3.8 Flash streams ~2× faster (116 vs 80 output tokens per second, Artificial Analysis 7-day median).
Which is better for long documents?
Gemini 3.8 Flash scores higher on long-context benchmarks (67 vs 64). Both accept 1,048,576 tokens of context.
More questions (3)
Which has lower latency?
Gemini 3.8 Flash answers first: 14.8 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 DeepSeek V4 Pro vs Gemini 3.8 Flash?
Across the 21 benchmarks both have, DeepSeek V4 Pro leads 2, Gemini 3.8 Flash leads 19. Every overall index that rates both puts Gemini 3.8 Flash ahead.
When was Gemini 3.8 Flash released?
Gemini 3.8 Flash was released on 2 Sep 2026.