BenchLeader

Qwen3.8 27B vs Qwen3.8 Max

Full profiles: Qwen3.8 27B (Alibaba) and Qwen3.8 Max (Alibaba): every benchmark, price and rank.

Updated · Data from 14 sources

As of 16 Nov 2026, Qwen3.8 27B and Qwen3.8 Max are neck and neck on quality (55.1 vs 58.7 on the BenchLeader Index, inside the ±6.6 margin), and Qwen3.8 27B is 2.8× cheaper at $1.09 vs $3.00 per 1M tokens blended. Choose 27B for reasoning (58 vs 52) and fast first answers (0.7 s vs 119.6 s); choose Max for agents & tools (59 vs 44), maths (65 vs 55), and human preference (68 vs 61).

Key facts (as of 16 Nov 2026)

  • Qwen3.8 27B scores 55.1 on the BenchLeader Index and Qwen3.8 Max scores 58.7 (BenchLeader, 16 Nov 2026).
  • Qwen3.8 27B costs $1.09 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 23 benchmarks both models have, Qwen3.8 Max leads 22 (BenchLeader, 16 Nov 2026).
  • On Arena WebDev, Qwen3.8 27B is rated 1590 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
CodingEither62 vs 62 category score; Arena WebDev 1590 vs 1676Either
Agents & tool useQwen3.8 Max59 vs 44 category score; APEX-Agents 63.3% vs 47.5%High
MathsQwen3.8 Max65 vs 55 category score; Arena Maths 1505 vs 1451High
ReasoningQwen3.8 27B58 vs 52 category score; Arena Hard Prompts 1459 vs 1507Medium
Images and other mediaQwen3.8 Max66 vs 61 category score; Arena Vision 1312 vs 1270Medium
Chat quality (human preference)Qwen3.8 Max68 vs 61 on Arena votesMedium
Lowest costQwen3.8 27B$1.06 vs $3.00 per 1M blended; agentic coding session $0.037 vs $0.144–
Fastest first answerQwen3.8 27B0.7 s (Qwen3.8 27B) 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

  • Qwen3.8 27B and Qwen3.8 Max (max) are neck and neck on quality (55.1 vs 58.7, within the ±6.6 margin).
  • Qwen3.8 27B is stronger in reasoning.
  • Qwen3.8 Max (max) is stronger in agents & tools, human preference, maths, multimodal.
  • Qwen3.8 27B is 2.8× cheaper ($1.09 vs $3.00 per 1M blended).

At a glance

SpecQwen3.8 27BQwen3.8 Max
DeveloperAlibabaAlibaba
Released14 Aug 20263 Aug 2026
Context window (tokens)262,1441,000,000
Compared atQwen3.8 27BQwen3.8 Max (max)
Open weightsYesNo
Price in / out / cached input per 1M$0.420 / $3.00 / $0.085$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.

CategoryQwen3.8 27BQwen3.8 Max (max)Leader
Reasoning57.651.5Qwen3.8 27B
Coding61.661.6level
Agents & tools44.258.5Qwen3.8 Max
Maths54.665.2Qwen3.8 Max
Knowledge–58.0not comparable
Human preference61.367.5Qwen3.8 Max
Multimodal60.765.7Qwen3.8 Max
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 avg75ReasoningCodingAgents & toolsMathsMultimodalHumanpreferenceQwen3.8 27B, Reasoning: 57.6Qwen3.8 27B, Coding: 61.6Qwen3.8 27B, Agents & tools: 44.2Qwen3.8 27B, Maths: 54.6Qwen3.8 27B, Multimodal: 60.7Qwen3.8 27B, Human preference: 61.3Qwen3.8 Max, Reasoning: 51.5Qwen3.8 Max, Coding: 61.6Qwen3.8 Max, Agents & tools: 58.5Qwen3.8 Max, Maths: 65.2Qwen3.8 Max, Multimodal: 65.7Qwen3.8 Max, Human preference: 67.5

Not drawn, measured for fewer than two of these models: Knowledge, Instruction following, Long context.

Model, click to add or removeIndex

Where sources disagree

Across the 23 benchmarks both have, Qwen3.8 27B leads 1, Qwen3.8 Max leads 22. Every overall index that rates both puts Qwen3.8 Max ahead.

Overall indexQwen3.8 27BQwen3.8 Max (max)Leader
BenchLeader Index55.158.7neck and neck
AA Intelligence Index v5.0.110.0†33.6Qwen3.8 Max
Arena Text14341484Qwen3.8 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.

MeasureQwen3.8 27BQwen3.8 Max (max)
Blended price per 1M tokens$1.06$3.00
Input / output per 1M$0.420 / $3.00$2.00 / $6.00
Cached input per 1M$0.085$0.250
Price per task: chat reply$0.0011$0.0026
Price per task: summarize a 30-page report$0.0068$0.028
Price per task: code edit$0.0070$0.021
Price per task: agentic coding session$0.037$0.144
Price per task: structured extraction$0.0014$0.0052
Agentic coding session, input mostly cached$0.022$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 speed73 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. Qwen3.8 27B scores 55.1 and takes 1 s to its first answer token; Qwen3.8 27B (xhigh) scores 52.2 in 109 s, Qwen3.8 27B (medium) scores 50.2 in 24 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

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.

MetricPublisherQwen3.8 27BQwen3.8 Max (max)
BenchLeader IndexBenchLeader55.158.7
Agents & tools scoreBenchLeader44.258.5
Coding scoreBenchLeader61.661.6
Human preference scoreBenchLeader61.367.5
Maths scoreBenchLeader54.665.2
Multimodal scoreBenchLeader60.765.7
Reasoning scoreBenchLeader57.651.5
Composite scoreBenchLeader–45.0
Knowledge scoreBenchLeader–58.0
Long context scoreBenchLeader–64.8
Blended price $/MBenchLeader$1.06$3.00
Output speedBenchLeader73 tok/s59 tok/s
Time to first answerBenchLeader0.7 s119.6 s
Context windowBenchLeader262,144 tokens1,000,000 tokens
Reasoning
GPQA DiamondEpoch AI Benchmarking Hub–92.7%†
SimpleBenchSimpleBench60.2%–
Arena Hard PromptsArena (LMArena)14591507
LiveBench 2.0 ReasoningLiveBench80.0%†88.2%
LiveBench 1.0 ReasoningLiveBench80.0%88.2%
GPQA Diamond (AA)Artificial Analysis90.5%†92.8%
Humanity's Last Exam (AA)Artificial Analysis33.9%†43.1%
HLE-Diamond (AA)Artificial Analysis15.3%†24.0%
GPQA Diamond (Vals)Vals AI88.9%†93.7%
ARC-AGI-1ARC Prize87.5%†–
ARC-AGI-2ARC Prize42.4%†–
CritPtArtificial Analysis5.4%†20.0%
CritPt v2 hard subset (AA)Artificial Analysis4.0%†29.5%
MysteryMechanismVals AI–23.9%
Arena ExpertArena (LMArena)14731514
Chess PuzzlesEpoch AI Benchmarking Hub–46.0%†
Mystery Game PuzzlesEpoch AI Benchmarking Hub–38.0%†
LMCAEpoch AI Benchmarking Hub41.4%46.2%†
DTBenchEpoch AI Benchmarking Hub88.0%92.0%†
Coding
SciCodeSciCode46.6%†53.2%
Arena CodingArena (LMArena)14961523
Arena WebDevArena (LMArena)15901676
LiveBench 2.0 CodingLiveBench69.3%†60.9%
LiveBench 1.0 CodingLiveBench75.7%72.9%
LiveCodeBenchVals AI84.0%†87.8%
Show all 147 benchmarks
MetricPublisherQwen3.8 27BQwen3.8 Max (max)
Coding
IOIVals AI39.1%†68.9%
SWE-bench (Vals)Vals AI86.0%†85.6%
Code MigrationVals AI14.2%†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 AI64.8%†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%†
Surface Evolver BenchEpoch AI Benchmarking Hub45.0%†–
DeepSWE v1.1Epoch AI Benchmarking Hub–57.5%†
ALE-BenchEpoch AI Benchmarking Hub–1679.1†
FrontierSWEEpoch AI Benchmarking Hub–17.8%†
Terminal-Bench 4.0 (AA)Artificial Analysis5.6%†38.9%
Terminal-Bench 2.1 (AA)Artificial Analysis79.8%†88.8%
Startrise FrontendStartrise–71.3%
GameBench 2 OverallDuelLab GameBench 2–54.2†
GameBench 2 by reasoning settingDuelLab GameBench 243.8†69.6†
Design Arena WebsiteDesign Arena–1269
FrontierCode 1.1 (AA)Artificial Analysis34.2%†43.2%
FrontierCode 2.0 (AA)Artificial Analysis14.9%†23.4%
Agents & tools
Terminal-BenchTerminal-Bench–27.0%
APEX-AgentsMercor47.5%63.3%
Arena AgentArena (LMArena)-4.80.1
LiveBench Long AgenticLiveBench15.9%†26.6%
LiveBench BrowserLiveBench24.2%†49.4%
LiveBench Agentic CodingLiveBench61.4%64.7%
Terminal-Bench 2.1 (Vals)Vals AI58.4%†67.4%
τ³-bench Bankingτ²-bench–55.1%†
GDPval-AA v2.1, normalised scoreArtificial Analysis46.0%†58.3%
GDPval-AA v2.1Artificial Analysis1426†1674
τ³-Banking (AA)Artificial Analysis48.0%†51.3%
ITBench SRE (AA)Artificial Analysis–40.3%
Analyst Agent (AA)Artificial Analysis–45.0%
APEX-Agents (AA)Artificial Analysis–42.4%
CyberBenchVals AI67.8%†28.6%
Finance Agent v2Vals AI48.5%†50.6%
Harvey's Legal Agent Benchmark v1Vals AI11.3%†10.4%
Legal Research BenchVals AI36.1%†47.6%
SkillsBenchVals AI38.1%†42.0%
Tax Agent BenchVals AI30.9%†32.1%
Terminal-Bench 4.0 (Vals)Vals AI–34.3%
Terminal-Bench ScienceVals AI0.0%†12.9%
Arena SearchArena (LMArena)–1227
Arena Agent Task CompletionArena (LMArena)-2.74.5
AutomationBenchArtificial Analysis48.2%†56.2%
GDP.pdf (AA)Artificial Analysis16.4%†22.8%
EnterpriseOps-GymArtificial Analysis44.2%†47.6%
AA-Briefcase v1.1Artificial Analysis1391†1604
GDPxGDPx by SemiAnalysis–3.5%
Design Arena AgenticDesign Arena–1264
Terminal-Bench-Science 0.1 (AA)Artificial Analysis0.5%†11.9%
CWE-Bench (AA)Artificial Analysis26.7%†–
DeepSec-Bench (AA)Artificial Analysis11.5%†–
CyberGym end-to-end (AA)Artificial Analysis0.0%†–
MKA (AA)Artificial Analysis–37.4%
FACTS SearchGoogle–85.0%
Maths
FrontierMath Tiers 1–3Epoch AI Benchmarking Hub–74.7%†
FrontierMath Tier 4Epoch AI Benchmarking Hub–46.3%†
OTIS Mock AIMEEpoch AI Benchmarking Hub–100.0%†
ProofBenchVals AI16.0%58.0%
LiveBench MathematicsLiveBench86.2%91.3%
Arena MathsArena (LMArena)14511505
Knowledge
SimpleQA VerifiedEpoch AI Benchmarking Hub–47.3%†
LiveBench 2.0 Data AnalysisLiveBench76.6%†78.4%
LiveBench ResearchLiveBench61.3%†72.8%
LiveBench 1.0 Data AnalysisLiveBench76.6%78.4%
AA-Omniscience v1.1Artificial Analysis-6.4†10.5
MMLU-ProVals AI84.3%†88.6%
LegalBenchVals AI82.4%†83.6%
CorpFinVals AI–65.8%
TaxEvalVals AI70.8%†75.5%
Excel Modeling BenchmarkVals AI59.7%†60.1%
Excel Modeling Benchmark v1.1Vals AI–66.0%
MedCodeVals AI28.7%†40.7%
MedScribeVals AI83.8%†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 Analysis17.3%†29.5%
AA-Omniscience v1.1: non-hallucinationArtificial Analysis71.4%†72.9%
FACTS ParametricGoogle–30.8%
Instruction following
LiveBench LanguageLiveBench74.3%79.7%
LiveBench Instruction FollowingLiveBench72.7%74.1%
Arena Instruction FollowingArena (LMArena)14191478
Human preference
Arena TextArena (LMArena)14341484
EQ-Bench Creative Writing v3EQ-Bench1671–
Arena Creative WritingArena (LMArena)13591472
Arena Multi-turnArena (LMArena)14361495
Arena Longer QueriesArena (LMArena)14431496
Multimodal
Arena VisionArena (LMArena)12701312
LiveBench DesignLiveBench72.4%†76.6%
MMMU-ProArtificial Analysis76.3%†82.8%
MMMU-Pro (Vals)Vals AI83.9%†88.0%
MortgageTaxVals AI64.9%†64.0%
SAGEVals AI–81.8%
Humanity's Sixth SenseScale AI SEAL–33.0%
GeoBenchGeoBench38.0%–
Furniture AssemblyEpoch AI Benchmarking Hub–20.0%†
VIALS-AAHandshake AI–31.7%
Code2Video-AAArtificial Analysis949†1021
AA-3DArtificial Analysis–954
FACTS MultimodalGoogle–51.5%
Long context
AA-LCR v1.1Artificial Analysis82.0%†80.3%
MLCRArtificial Analysis22.8%†26.1%
FACTS GroundingGoogle–73.8%
Composite
Epoch Capabilities IndexEpoch AI Benchmarking Hub152.6157.6
LiveBench 2.0LiveBench54.0%†60.9%
LiveBench 1.0LiveBench75.3%78.5%
AA Intelligence Index v5.0.1Artificial Analysis20.1†33.6
AA Intelligence Index v4.3.2Artificial Analysis33.7†45.4
AA Progress Index v1Artificial Analysis303.1†323.7
Vals RSI IndexVals AI–21.8%
Vals Multimodal IndexVals AI–65.4%
AA Cyber IndexArtificial Analysis12.7†–
Safety & honesty
DistressBenchScale AI SEAL–77.6%
Metadata
AA Openness IndexArtificial Analysis38.9†–

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 27B: BenchLeader Index 53.2 → 55.1.
  • Qwen3.8 27B: Qwen3.8 27B cached input price moved from $0.10/M to $0.085/M (15% down).
  • Qwen3.8 Max (max): 1 score revision.
  • Qwen3.8 27B: Qwen3.8 27B input price moved from $0.45/M to $0.42/M (7% down).
  • Qwen3.8 Max (max): 1 score revision.
  • Qwen3.8 27B: BenchLeader Index 54.0 → 53.2.
  • Qwen3.8 27B: 2 new results (SimpleBench, GeoBench).
  • Qwen3.8 27B: 5 score revisions.
  • Qwen3.8 Max (max): 6 score revisions.
  • Qwen3.8 Max (max): 1 score revision.

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

Qwen3.8 27B vs Qwen3.8 Max: questions

Is Qwen3.8 27B better than Qwen3.8 Max?
Qwen3.8 27B and Qwen3.8 Max (max) are neck and neck on quality (55.1 vs 58.7, within the ±6.6 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.
Qwen3.8 27B vs Qwen3.8 Max: which is cheaper, and will that change?
Qwen3.8 27B is cheaper today: $1.09 against $3.00 per million tokens, blended at three input tokens per output token. No price change has been announced for either.
Is Qwen3.8 27B or Qwen3.8 Max better for coding?
They are level in coding (62 vs 62 on the category index). On Arena WebDev, the top shared coding benchmark, it is 1590 for Qwen3.8 27B against 1676 for Qwen3.8 Max.
Which is faster, Qwen3.8 27B or Qwen3.8 Max?
Qwen3.8 27B streams faster (73 vs 59 output tokens per second, Artificial Analysis 7-day median).
Is Qwen3.8 27B or Qwen3.8 Max better for agents and tool use?
Qwen3.8 Max scores higher in agentic tasks (59 vs 44 on the category index, where 50 is average). On APEX-Agents, the top shared agents & tools benchmark, it is 47.5% for Qwen3.8 27B against 63.3% for Qwen3.8 Max.
More questions (4)
Which is better for long documents?
Qwen3.8 Max accepts more context: 1,000,000 against 262,144 tokens for Qwen3.8 27B.
Which has lower latency?
Qwen3.8 27B 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 Qwen3.8 27B vs Qwen3.8 Max?
Across the 23 benchmarks both have, Qwen3.8 27B leads 1, Qwen3.8 Max leads 22. Every overall index that rates both puts Qwen3.8 Max ahead.
When was Qwen3.8 27B released?
Qwen3.8 27B was released on 14 Aug 2026.