Atlas

Compare Models

Kimi K3.1 Pro vs Qwen4-Max

Qwen4-Max leads by 0.4 capability points (≈95% interval −2.1 to +2.8): statistically tied · P(higher) ≈ 61%

Wins 26 of 40 shared benchmarks (10 lost, 4 tied)

Capability vs. relative cost
Kimi K3.1 ProQwen4-MaxClosed weightsOpen weights

Benchmark wins10 – 26 · 40 shared, 4 tied

Comparing matching setups only (38 benchmarks excluded) · suites counted once
← Kimi K3.1 ProQwen4-Max →
Data: Vals AI, Surge AI, Arena and 5 othersatlas.village.dev

Margins

Each benchmark scaled 0–100% (its declared range, else the field’s range); Δ is Qwen4-Max minus Kimi K3.1 Pro.

DAYJOB: Healthcare (October 2026 rescoring)Health & medicine8.824.4+15.6
Terminal-Bench 4.0Software › Software engineering63.648.5−15.2
Legal Research Bench Weighted PassProfessional › Legal71.484.7+13.3
LLM Debate BenchmarkLanguage & writing › Language & wordplay16011679+11.1
GDP.xlsxProfessional › General knowledge work14.023.7+9.7
HLAB-Vals-1.1Professional › Legal11.721.3+9.6
Legal Research Bench Business/CommercialProfessional › Legal27.937.2+9.3
Legal Research Bench ImmigrationProfessional › Legal18.227.3+9.1
DAYJOB: AccountingProfessional › Finance, accounting & tax13.622.4+8.8
LMArena Code Arena WebDevSoftware › Frontend & design16841750+8.7
DAYJOB: Finance (October 2026 rescoring)Professional › Finance, accounting & tax17.025.3+8.3
DAYJOB: TaxProfessional › Finance, accounting & tax1.68.8+7.2
LMArena Agent Arena Net Improvement (Steering Burden)Tool & computer use › General agents3.97.3+6.6
Tax Agent BenchProfessional › Finance, accounting & tax63.970.0+6.2
Vals Finance IndexProfessional › Finance, accounting & tax56.062.0+6.0
Humanity's Last Exam DiamondKnowledge & reasoning › General knowledge39.444.9+5.5
Legal Research Bench CriminalProfessional › Legal28.133.3+5.3
Finance Agent v2.1Professional › Finance, accounting & tax61.566.7+5.2
DAYJOB: Law (November 2026 release)Professional › Legal22.017.5−4.5
LiveBench 2.0 (2026-11-06 release)Composite indexes68.773.1+4.4
Legal Research Bench Civil LitigationProfessional › Legal29.633.3+3.7
LMArena Agent Arena Confirmed SuccessTool & computer use › General agents1.98.8+3.4
LMArena Agent Arena Praise vs ComplaintTool & computer use › General agents6.713.0+3.1
GDP.pdf (November 2026 rescoring)Professional › General knowledge work46.643.6−3.0
DualEntry Accounting AI BenchmarkProfessional › Finance, accounting & tax86.183.2−2.9
EnterpriseBench: CoreCraftTool & computer use › Customer service & operations66.168.7+2.6
Code Migration v1.1 (Vals Index v2.2 subset)Software › Software engineering59.757.3−2.4
LMArena Agent Arena Steering BurdenTool & computer use › General agents4.58.0+1.8
Vibe Code Bench v1.1Software › Frontend & design84.986.7+1.8
GDPval-AA v2.1Professional › General knowledge work18461809−1.6
LMArena Agent Arena Bash ReliabilityTool & computer use › General agents6.23.4−1.4
sudo L7 (October 2026 rescoring)Software › Software engineering16.715.6−1.1
Excel Modeling Benchmark v1.1Professional › Finance, accounting & tax70.070.8+0.8
AutomationBench-AA 1.0.6Tool & computer use › Customer service & operations73.372.6−0.7
AA-Briefcase v1.1 EloProfessional › General knowledge work16781686+0.4
LMArena Agent Arena Tool HallucinationTool & computer use › Tool calling1.21.20.0
Legal Research Bench Administrative/RegulatoryProfessional › Legal56.156.10.0
Legal Research Bench Constitutional/Civil RightsProfessional › Legal31.631.60.0
Legal Research Bench FamilyProfessional › Legal13.613.60.0
Legal Research Bench HealthProfessional › Legal73.373.30.0
Data: Vals AI, Surge AI, Arena and 5 othersatlas.village.dev
Win counts use each benchmark’s declared score direction. Represented suite slices and the derived Capability score are excluded from headline totals. Missing or incompatible setups are excluded, never counted as losses.
Loading Atlas data…