Atlas

Compare Models

Gemini 4 Hydrogen vs Kimi K3.1 Pro

Gemini 4 Hydrogen leads by 2.4 capability points (≈95% interval −0.1 to +4.9): statistically tied · P(higher) ≈ 97%

Wins 48 of 61 shared benchmarks (11 lost, 2 tied)

Capability vs. relative cost
Gemini 4 HydrogenKimi K3.1 ProClosed weightsOpen weights

Benchmark wins48 – 11 · 61 shared, 2 tied

Comparing matching setups only (35 benchmarks excluded) · suites counted once
← Gemini 4 HydrogenKimi K3.1 Pro →
Data: Vals AI, Epoch AI, Arena and 15 othersatlas.village.dev

Margins

Each benchmark scaled 0–100% (its declared range, else the field’s range); Δ is Kimi K3.1 Pro minus Gemini 4 Hydrogen.

EBR-bench v4 (card ban, single agent)Games70.537.1−33.3
Humanity's Sixth SensePerception & media › Spatial & visual reasoning63.433.2−30.2
Legal Research Bench Constitutional/Civil RightsProfessional › Legal60.531.6−28.9
Legal Research Bench ImmigrationProfessional › Legal45.518.2−27.3
GDP.xlsxProfessional › General knowledge work40.914.0−26.9
Ramp SWE-Bench v2 (Bash harness)Software › Software engineering70.845.4−25.4
Legal Research Bench CriminalProfessional › Legal52.628.1−24.6
Legal Research Bench Business/CommercialProfessional › Legal51.227.9−23.3
Mystery Game PuzzlesGames65.042.0−23.0
Code Migration v1.1 (Vals Index v2.2 subset)Software › Software engineering82.759.7−23.0
CWE-bench v2 Pass@1Software › Cybersecurity68.145.3−22.8
Legal Research Bench Civil LitigationProfessional › Legal51.929.6−22.2
FrontierMath Tier 4 (v2) - PrivateMath87.865.9−22.0
DAYJOB: Finance (October 2026 rescoring)Professional › Finance, accounting & tax38.517.0−21.5
FrontierCode 2.0 - Pass RateSoftware › Software engineering43.222.0−21.2
DAYJOB: AccountingProfessional › Finance, accounting & tax33.613.6−20.0
Humanity's Last Exam DiamondKnowledge & reasoning › General knowledge58.839.4−19.4
Legal Research Bench Weighted PassProfessional › Legal90.871.4−19.3
FrontierCode 2.0 - ScoreSoftware › Software engineering38.919.9−19.0
LMArena Agent Arena Net Improvement (Steering Burden)Tool & computer use › General agents13.53.9−18.8
DAYJOB: Healthcare (October 2026 rescoring)Health & medicine24.08.8−15.2
Chess Puzzles (task version 2.0)Games60.045.0−15.0
sudo L7 (October 2026 rescoring)Software › Software engineering31.716.7−15.0
Tax Agent BenchProfessional › Finance, accounting & tax78.463.9−14.6
Legal Research Bench FamilyProfessional › Legal27.313.6−13.6
Legal Research Bench Administrative/RegulatoryProfessional › Legal69.756.1−13.6
DAYJOB: Law (November 2026 release)Professional › Legal9.722.0+12.3
LMArena Agent Arena Praise vs ComplaintTool & computer use › General agents29.36.7−11.3
Vals Finance IndexProfessional › Finance, accounting & tax66.156.0−10.1
FrontierMath Tiers 1-3 (v2) - PrivateMath88.178.2−9.8
Terminal-Bench 4.0Software › Software engineering54.063.6+9.6
GBENCH One-shot Coding GScoreGames66.858.0−8.8
SimpleQA Verified (Epoch anti-abstention)Knowledge & reasoning › General knowledge70.962.2−8.7
LiveBench 2.0 (2026-11-06 release)Composite indexes77.068.7−8.3
SpireClimbGames7.70.0−7.7
GDP.pdf (November 2026 rescoring)Professional › General knowledge work54.246.6−7.6
LMArena Agent Arena Confirmed SuccessTool & computer use › General agents15.71.9−6.9
Legal Research Bench HealthProfessional › Legal80.073.3−6.7
LMArena Agent Arena Steering BurdenTool & computer use › General agents17.44.5−6.5
micro1 PharmaBenchProfessional › General knowledge work11.417.8+6.4
HiL-Bench (Human-in-Loop Benchmark)Tool & computer use › General agents55.361.3+6.0
AA-Briefcase v1.1 EloProfessional › General knowledge work15691678+6.0
Excel Modeling Benchmark v1.1Professional › Finance, accounting & tax75.870.0−5.8
LMArena Agent Arena Bash ReliabilityTool & computer use › General agents-4.56.2+5.4
Epoch Capabilities Index (ECI)Composite indexes167161−4.5
Artificial Analysis Coding Agent Index v2.1Software › Software engineering52.348.6−3.7
Finance Agent v2.1Professional › Finance, accounting & tax65.161.5−3.6
Bug Hunt BenchSoftware › Software engineering31.134.3+3.2
DualEntry Accounting AI BenchmarkProfessional › Finance, accounting & tax83.286.1+2.9
Vibe Code Bench v1.1Software › Frontend & design87.884.9−2.9
LMArena Code Arena WebDevSoftware › Frontend & design17031684−2.4
DAYJOB: TaxProfessional › Finance, accounting & tax3.61.6−2.0
Furniture Assembly BenchmarkPerception & media › Spatial & visual reasoning57.555.8−1.7
BigFinanceBench private holdout (final-answer accuracy)Professional › Finance, accounting & tax49.448.2−1.2
AutomationBench-AA 1.0.6Tool & computer use › Customer service & operations74.473.3−1.1
BuseyBench SVGPerception & media › Generation6.96.9+0.8
LMArena Text Arena EloLanguage & writing › General assistance14901487−0.6
HLAB-Vals-1.1Professional › Legal11.311.7+0.4
LMArena Agent Arena Tool HallucinationTool & computer use › Tool calling0.71.2+0.2
Agents' Last Exam v1.1 (ALE v1.1)Professional › General knowledge work43.443.40.0
OTIS Mock AIME 2024-2025Math100.0100.00.0
Data: Vals AI, Epoch AI, Arena and 15 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…