BenchLeader

DeepSeek V4 Flash vs Qwen3.8 Max

Full profiles: DeepSeek V4 Flash (DeepSeek) and Qwen3.8 Max (Alibaba): every benchmark, price and rank.

Updated · Data from 15 sources

As of 16 Nov 2026, DeepSeek V4 Flash and Qwen3.8 Max are neck and neck on quality (57.2 vs 58.7 on the BenchLeader Index, inside the ±5.2 margin), and DeepSeek V4 Flash is 11× cheaper at $0.262 vs $3.00 per 1M tokens blended. Choose DeepSeek V4 Flash for reasoning (59 vs 52) and fast first answers (0.7 s vs 119.6 s); choose Qwen3.8 Max for coding (62 vs 55), human preference (68 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 Qwen3.8 Max scores 58.7 (BenchLeader, 16 Nov 2026).
  • DeepSeek V4 Flash costs $0.262 per 1M tokens blended (list price, 16 Nov 2026).
  • Qwen3.8 Max costs $3.00 per 1M tokens blended (list price, 16 Nov 2026).
  • Of the 32 benchmarks both models have, Qwen3.8 Max leads 25 (BenchLeader, 16 Nov 2026).
  • On Arena WebDev, DeepSeek V4 Flash is rated 1582 and Qwen3.8 Max 1676 (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
CodingQwen3.8 Max62 vs 55 category score; Arena WebDev 1676 vs 1582Medium
MathsQwen3.8 Max65 vs 62 category score; Arena Maths 1505 vs 1441Medium
ReasoningDeepSeek V4 Flash59 vs 52 category score; Arena Hard Prompts 1459 vs 1507High
KnowledgeQwen3.8 Max58 vs 55 category score; MMLU-Pro 88.6% vs 86.2%Medium
Chat quality (human preference)Qwen3.8 Max68 vs 62 on Arena votesMedium
Lowest costDeepSeek V4 Flash$0.262 vs $3.00 per 1M blended; agentic coding session $0.011 vs $0.144–
Fastest first answerDeepSeek V4 Flash0.7 s (DeepSeek V4 Flash (high)) vs 119.6 s (Qwen3.8 Max (max)) 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 Qwen3.8 Max (max) are neck and neck on quality (57.2 vs 58.7, within the ±5.2 margin).
  • DeepSeek V4 Flash (high) is stronger in reasoning.
  • Qwen3.8 Max (max) is stronger in coding, human preference, knowledge, maths.
  • DeepSeek V4 Flash (high) is 11× cheaper ($0.262 vs $3.00 per 1M blended).
  • DeepSeek V4 Flash (high) streams 1.4× faster (82 vs 59 tokens per second).

At a glance

SpecDeepSeek V4 FlashQwen3.8 Max
DeveloperDeepSeekAlibaba
Released31 Jul 20263 Aug 2026
Context window (tokens)1,048,5761,000,000
Compared atDeepSeek V4 Flash (high)Qwen3.8 Max (max)
Open weightsYesNo
Price in / out / cached input per 1M$0.150 / $0.600 / $0.0030$2.00 / $6.00 / $0.250

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)Qwen3.8 Max (max)Leader
Reasoning59.251.5DeepSeek V4 Flash
Coding55.061.6Qwen3.8 Max
Agents & tools–58.5not comparable
Maths61.965.2Qwen3.8 Max
Knowledge54.858.0Qwen3.8 Max
Human preference62.067.5Qwen3.8 Max
Multimodal–65.7not comparable
Long context–64.8not comparable
Composite–45.0not 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: 62.0Qwen3.8 Max, Reasoning: 51.5Qwen3.8 Max, Coding: 61.6Qwen3.8 Max, Maths: 65.2Qwen3.8 Max, Knowledge: 58.0Qwen3.8 Max, Human preference: 67.5

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 32 benchmarks both have, DeepSeek V4 Flash leads 6, Qwen3.8 Max leads 25 and 1 is a tie. Every overall index that rates both puts Qwen3.8 Max ahead.

Overall indexDeepSeek V4 Flash (high)Qwen3.8 Max (max)Leader
BenchLeader Index57.258.7neck and neck
Arena Text14391484Qwen3.8 Max

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)Qwen3.8 Max (max)
Blended price per 1M tokens$0.262$3.00
Input / output per 1M$0.150 / $0.600$2.00 / $6.00
Cached input per 1M$0.0030$0.250
Price per task: chat reply$0.0002$0.0026
Price per task: summarize a 30-page report$0.0022$0.028
Price per task: code edit$0.0018$0.021
Price per task: agentic coding session$0.011$0.144
Price per task: structured extraction$0.0004$0.0052
Agentic coding session, input mostly cached$0.0048$0.065
Cost per benchmark task (AA Index)–$10.02
Output tokens per task (AA Index)–171k
Of which reasoning–120k
Time per task (AA Index)–48 min
Output speed82 tok/s59 tok/s
Time to first answer0.7 s119.6 s
Time to a complete answer, last measured 13 Nov 2026–294.1 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. Qwen3.8 Max (max) scores 58.7 and takes 120 s to its first answer token; Qwen3.8 Max (xhigh) scores 56.6 in 3 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

150 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)Qwen3.8 Max (max)
BenchLeader IndexBenchLeader57.258.7
Coding scoreBenchLeader55.061.6
Human preference scoreBenchLeader62.067.5
Knowledge scoreBenchLeader54.858.0
Maths scoreBenchLeader61.965.2
Reasoning scoreBenchLeader59.251.5
Agents & tools scoreBenchLeader–58.5
Composite scoreBenchLeader–45.0
Long context scoreBenchLeader–64.8
Multimodal scoreBenchLeader–65.7
Blended price $/MBenchLeader$0.262$3.00
Output speedBenchLeader82 tok/s59 tok/s
Time to first answerBenchLeader0.7 s119.6 s
Context windowBenchLeader1,048,576 tokens1,000,000 tokens
Reasoning
GPQA DiamondEpoch AI Benchmarking Hub91.0%†92.7%†
SimpleBenchSimpleBench61.1%†–
Arena Hard PromptsArena (LMArena)14591507
LiveBench 2.0 ReasoningLiveBench–88.2%
LiveBench 1.0 ReasoningLiveBench86.6%†88.2%
GPQA Diamond (AA)Artificial Analysis90.8%†92.8%
Humanity's Last Exam (AA)Artificial Analysis38.5%†43.1%
HLE-Diamond (AA)Artificial Analysis–24.0%
GPQA Diamond (Vals)Vals AI89.9%93.7%
Kagi LLM BenchmarkKagi LLM Benchmark52.2%†–
ARC-AGI-1ARC Prize87.0%–
ARC-AGI-2ARC Prize56.0%–
CritPtArtificial Analysis16.6%†20.0%
CritPt v2 hard subset (AA)Artificial Analysis–29.5%
MysteryMechanismVals AI–23.9%
Arena ExpertArena (LMArena)14671514
Chess PuzzlesEpoch AI Benchmarking Hub32.0%†46.0%†
Mystery Game PuzzlesEpoch AI Benchmarking Hub34.0%†38.0%†
LMCAEpoch AI Benchmarking Hub41.7%†46.2%†
DTBenchEpoch AI Benchmarking Hub90.9%†92.0%†
Coding
SciCodeSciCode42.0%53.2%
WeirdMLWeirdML57.0%–
FrontierCodeCognition18.8%†–
Arena CodingArena (LMArena)14801523
Arena WebDevArena (LMArena)15821676
Show all 150 benchmarks
MetricPublisherDeepSeek V4 Flash (high)Qwen3.8 Max (max)
Coding
LiveBench 2.0 CodingLiveBench–60.9%
LiveBench 1.0 CodingLiveBench75.0%†72.9%
SciCode (AA)Artificial Analysis50.4%†53.2%
LiveCodeBenchVals AI87.3%87.8%
IOIVals AI32.7%68.9%
SWE-bench (Vals)Vals AI88.8%85.6%
Code MigrationVals AI38.6%24.0%
ProgramBenchVals AI0.0%0.0%
SWE-bench MultimodalVals AI–42.5%
Vibe Code Bench 1-100Vals AI–12.8%
Vibe Code Bench v1.1Vals AI74.7%64.7%
SWE Atlas v1.1Scale AI SEAL–64.4%
SWE Atlas v1.1: Codebase QnAScale AI SEAL–81.5%
SWE Atlas v1.1: Test WritingScale AI SEAL–58.9%
SWE Atlas v1.1: RefactoringScale AI SEAL–52.9%
MirrorCodeEpoch AI Benchmarking Hub–52.2%†
DeepSWE v1.1Epoch AI Benchmarking Hub–57.5%†
ALE-BenchEpoch AI Benchmarking Hub678.21679.1†
FrontierSWEEpoch AI Benchmarking Hub–17.8%†
Terminal-Bench 4.0 (AA)Artificial Analysis12.1%†38.9%
Terminal-Bench 2.1 (AA)Artificial Analysis78.7%†88.8%
Startrise FrontendStartrise–71.3%
GameBench 2 OverallDuelLab GameBench 234.1†54.2†
GameBench 2 by reasoning settingDuelLab GameBench 243.469.6†
Design Arena WebsiteDesign Arena1199†1269
FrontierCode 1.1 (AA)Artificial Analysis–43.2%
FrontierCode 2.0 (AA)Artificial Analysis–23.4%
Agents & tools
Terminal-BenchTerminal-Bench–27.0%
APEX-AgentsMercor–63.3%
Arena AgentArena (LMArena)–0.1
LiveBench Long AgenticLiveBench–26.6%
LiveBench BrowserLiveBench–49.4%
LiveBench Agentic CodingLiveBench46.8%†64.7%
Terminal-Bench 2.1 (Vals)Vals AI67.0%†67.4%
HiL-BenchScale AI SEAL47.8%†–
τ³-bench Bankingτ²-bench–55.1%†
GDPval-AA v2.1, normalised scoreArtificial Analysis–58.3%
GDPval-AA v2.1Artificial Analysis1437†1674
τ³-Banking (AA)Artificial Analysis39.4%†51.3%
ITBench SRE (AA)Artificial Analysis–40.3%
Analyst Agent (AA)Artificial Analysis–45.0%
APEX-Agents (AA)Artificial Analysis–42.4%
CyberBenchVals AI72.1%28.6%
Finance Agent v2Vals AI49.5%50.6%
Harvey's Legal Agent Benchmark v1Vals AI8.3%10.4%
Legal Research BenchVals AI30.3%47.6%
SkillsBenchVals AI50.7%42.0%
Tax Agent BenchVals AI28.1%32.1%
Terminal-Bench 4.0 (Vals)Vals AI18.7%34.3%
Terminal-Bench ScienceVals AI4.3%12.9%
Arena SearchArena (LMArena)–1227
Arena Agent Task CompletionArena (LMArena)–4.5
AutomationBenchArtificial Analysis54.0%†56.2%
GDP.pdf (AA)Artificial Analysis11.0%†22.8%
EnterpriseOps-GymArtificial Analysis–47.6%
AA-Briefcase v1.1Artificial Analysis1255†1604
GDPxGDPx by SemiAnalysis–3.5%
Design Arena AgenticDesign Arena–1264
Terminal-Bench-Science 0.1 (AA)Artificial Analysis–11.9%
MKA (AA)Artificial Analysis–37.4%
FACTS SearchGoogle–85.0%
Maths
FrontierMath Tiers 1–3Epoch AI Benchmarking Hub57.5%†74.7%†
FrontierMath Tier 4Epoch AI Benchmarking Hub24.4%†46.3%†
OTIS Mock AIMEEpoch AI Benchmarking Hub94.4%†100.0%†
ProofBenchVals AI56.0%†58.0%
LiveBench MathematicsLiveBench86.8%†91.3%
AIME 2026MathArena95.8%†–
HMMT February 2026MathArena93.9%†–
MathArena ApexMathArena27.1%†–
Arena MathsArena (LMArena)14411505
Knowledge
SimpleQA VerifiedEpoch AI Benchmarking Hub33.6%†47.3%†
LiveBench 2.0 Data AnalysisLiveBench–78.4%
LiveBench ResearchLiveBench–72.8%
LiveBench 1.0 Data AnalysisLiveBench79.3%†78.4%
AA-Omniscience v1.1Artificial Analysis–10.5
MMLU-ProVals AI86.2%88.6%
LegalBenchVals AI77.7%83.6%
CorpFinVals AI61.9%65.8%
TaxEvalVals AI70.7%75.5%
BioMysteryBenchVals AI64.4%–
Excel Modeling BenchmarkVals AI57.0%60.1%
Excel Modeling Benchmark v1.1Vals AI–66.0%
MedCodeVals AI41.4%40.7%
MedScribeVals AI80.4%85.0%
Public Benefits Bench v1.1Vals AI–67.1%
Access to Justice BenchmarkVals AI–76.4%
Financial Coaching BenchVals AI–75.6%
Immigration BenchVals AI–70.1%
GDP.pdfEpoch AI Benchmarking Hub–23.2%†
AA-Omniscience v1.1: accuracyArtificial Analysis–29.5%
AA-Omniscience v1.1: non-hallucinationArtificial Analysis–72.9%
FACTS ParametricGoogle–30.8%
Instruction following
LiveBench LanguageLiveBench79.2%†79.7%
LiveBench Instruction FollowingLiveBench65.5%†74.1%
Arena Instruction FollowingArena (LMArena)14341478
Human preference
Arena TextArena (LMArena)14391484
EQ-Bench Creative Writing v3EQ-Bench1559†–
Arena Creative WritingArena (LMArena)14111472
Arena Multi-turnArena (LMArena)14471495
Arena Longer QueriesArena (LMArena)14491496
Multimodal
Arena VisionArena (LMArena)–1312
LiveBench DesignLiveBench–76.6%
MMMU-ProArtificial Analysis–82.8%
MMMU-Pro (Vals)Vals AI–88.0%
MortgageTaxVals AI–64.0%
SAGEVals AI–81.8%
Humanity's Sixth SenseScale AI SEAL–33.0%
Furniture AssemblyEpoch AI Benchmarking Hub–20.0%†
VIALS-AAHandshake AI–31.7%
Code2Video-AAArtificial Analysis–1021
AA-3DArtificial Analysis–954
FACTS MultimodalGoogle–51.5%
Long context
AA-LCR v1.1Artificial Analysis79.7%†80.3%
MLCRArtificial Analysis–26.1%
FACTS GroundingGoogle–73.8%
Composite
Epoch Capabilities IndexEpoch AI Benchmarking Hub154.4†157.6
LiveBench 2.0LiveBench–60.9%
LiveBench 1.0LiveBench74.2%†78.5%
AA Intelligence Index v5.0.1Artificial Analysis–33.6
AA Intelligence Index v4.3.2Artificial Analysis34.3†45.4
AA Progress Index v1Artificial Analysis309.4†323.7
Vals RSI IndexVals AI–21.8%
Vals Multimodal IndexVals AI–65.4%
Safety & honesty
DistressBenchScale AI SEAL–77.6%
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

  • Qwen3.8 Max (max): 1 score revision.
  • DeepSeek V4 Flash (high): 1 score revision.
  • Qwen3.8 Max (max): 1 score revision.
  • Qwen3.8 Max (max): 6 score revisions.
  • Qwen3.8 Max (max): 1 score revision.
  • Qwen3.8 Max (max): 1 new result (FACTS Search); 1 score revision.
  • DeepSeek V4 Flash (high): 1 score revision.
  • Qwen3.8 Max (max): 1 score revision; cached input price moved from $0.22/M to $0.25/M (13% up).
  • DeepSeek V4 Flash (high): 1 new result (GameBench 2 by reasoning setting).
  • Qwen3.8 Max (max): cached input price moved from $0.25/M to $0.22/M (11% down).

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

DeepSeek V4 Flash vs Qwen3.8 Max: questions

Is DeepSeek V4 Flash better than Qwen3.8 Max?
DeepSeek V4 Flash (high) and Qwen3.8 Max (max) are neck and neck on quality (57.2 vs 58.7, within the ±5.2 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 Qwen3.8 Max: which is cheaper, and will that change?
DeepSeek V4 Flash is cheaper today: $0.262 against $3.00 per million tokens, blended at three input tokens per output token. No price change has been announced for either.
Is DeepSeek V4 Flash or Qwen3.8 Max better for coding?
Qwen3.8 Max scores higher in coding (62 vs 55 on the category index, where 50 is average). On Arena WebDev, the top shared coding benchmark, it is 1582 for DeepSeek V4 Flash against 1676 for Qwen3.8 Max.
Which is faster, DeepSeek V4 Flash or Qwen3.8 Max?
DeepSeek V4 Flash streams faster (82 vs 59 output tokens per second, Artificial Analysis 7-day median).
Which is better for long documents?
Both take about 1M tokens: DeepSeek V4 Flash 1,048,576, Qwen3.8 Max 1,000,000.
More questions (3)
Which has lower latency?
DeepSeek V4 Flash answers first: 0.7 s against 119.6 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 Qwen3.8 Max?
Across the 32 benchmarks both have, DeepSeek V4 Flash leads 6, Qwen3.8 Max leads 25 and 1 is a tie. Every overall index that rates both puts Qwen3.8 Max ahead.
When was Qwen3.8 Max released?
Qwen3.8 Max was released on 3 Aug 2026.