Atlas

Compare Models

Gemini 4 Hydrogen vs GPT-6.1 Sol

GPT-6.1 Sol leads by 0.5 capability points (≈95% interval −2.1 to +3.1): statistically tied · P(higher) ≈ 64%

Gemini 4 Hydrogen wins 38 of 73 shared benchmarks (31 lost, 4 tied)

Capability vs. relative cost
Gemini 4 HydrogenGPT-6.1 SolClosed weightsOpen weights

Benchmark wins38 – 31 · 73 shared, 4 tied

Comparing matching setups only (11 benchmarks excluded) · suites counted once
← Gemini 4 HydrogenGPT-6.1 Sol →
Data: Artificial Analysis, Vals AI, Epoch AI and 19 othersatlas.village.dev

Margins

Each benchmark scaled 0–100% (its declared range, else the field’s range); Δ is GPT-6.1 Sol minus Gemini 4 Hydrogen.

TxBench-Antibody-Development V1 (60 evaluations)Science › Life sciences60.627.8−32.8
Furniture Assembly BenchmarkPerception & media › Spatial & visual reasoning57.580.0+22.5
DAYJOB: Law (November 2026 release)Professional › Legal9.731.6+21.9
GDP.xlsxProfessional › General knowledge work40.922.0−18.9
Ramp SWE-Bench v2 (Bash harness)Software › Software engineering70.852.0−18.8
Humanity's Sixth SensePerception & media › Spatial & visual reasoning63.446.6−16.8
EBR-bench v4 (card ban, single agent)Games70.554.3−16.2
SpireClimbGames7.723.8+16.2
Tax Agent BenchProfessional › Finance, accounting & tax78.462.3−16.1
HANDBOOK.md Agents v1.1Tool & computer use › Customer service & operations44.628.6−16.0
Legal Research BenchProfessional › Legal53.838.5−15.4
Mystery Game PuzzlesGames65.080.0+15.0
Artificial Analysis Intelligence Index v5.0.1Composite indexes41.254.6+13.4
FrontierMath Tier 4 (v2) - PrivateMath87.8100.0+12.2
Medical Long Context Reasoning (MLCR-AA, October 2026 methodology)Health & medicine27.238.3+11.1
BuseyBench SVGPerception & media › Generation6.98.0+10.9
GBENCH One-shot Coding GScoreGames66.877.5+10.7
Vals Finance IndexProfessional › Finance, accounting & tax66.155.8−10.3
SimpleBenchKnowledge & reasoning › General knowledge91.782.9−8.8
Gray Swan IPI Attack Success (k=15)Safety & alignment0.08.6−8.6
DAYJOB: AccountingProfessional › Finance, accounting & tax33.625.2−8.4
Chess Puzzles (task version 2.0)Games60.068.0+8.0
TaxCalcBench TY25Professional › Finance, accounting & tax44.036.0−8.0
TxBench Antibody Discovery V1 (100 evaluations)Science › Life sciences57.749.7−8.0
AutomationBench-AA 1.0.6Tool & computer use › Customer service & operations74.466.6−7.9
Code Migration v1.1 (Vals Index v2.2 subset)Software › Software engineering82.775.1−7.5
HiL-Bench (Human-in-Loop Benchmark)Tool & computer use › General agents55.362.3+7.0
Professional Reasoning Benchmark - FinanceProfessional › Finance, accounting & tax52.346.2−6.1
FrontierMath Tiers 1-3 (v2) - PrivateMath88.193.7+5.6
Professional Reasoning Benchmark - LegalProfessional › Legal53.147.9−5.2
DAYJOB: Finance (October 2026 rescoring)Professional › Finance, accounting & tax38.533.5−5.0
HLAB-Vals-1.1Professional › Legal11.36.3−5.0
Vals Legal IndexProfessional › Legal73.668.7−4.9
LMArena Agent Arena Bash ReliabilityTool & computer use › General agents-4.55.1+4.8
Excel Modeling Benchmark v1.1Professional › Finance, accounting & tax75.871.1−4.7
Vals Index v2.2Composite indexes68.563.9−4.5
LMArena Agent Arena Net Improvement (Steering Burden)Tool & computer use › General agents13.511.3−4.3
LiveBench 2.0 (2026-11-06 release)Composite indexes77.081.2+4.2
Terminal-Bench 4.0Software › Software engineering54.058.2+4.1
ComplexConstraints (November 2026 rescoring)Language & writing › Instruction following73.770.1−3.6
AA-Briefcase v1.1 EloProfessional › General knowledge work15691504−3.5
Finance Agent v2.1Professional › Finance, accounting & tax65.161.8−3.3
SimpleQA Verified (Epoch anti-abstention)Knowledge & reasoning › General knowledge70.973.9+3.0
LMArena Agent Arena Steering BurdenTool & computer use › General agents17.411.8−2.8
FORTRESS Average Risk Score (ARS)Safety & alignment17.514.7+2.8
BigFinanceBench private holdout (final-answer accuracy)Professional › Finance, accounting & tax49.452.0+2.6
GDP.pdf (November 2026 rescoring)Professional › General knowledge work54.251.6−2.6
LMArena Agent Arena Praise vs ComplaintTool & computer use › General agents29.324.5−2.4
sudo L7 (October 2026 rescoring)Software › Software engineering31.733.9+2.2
LMArena Agent Arena Confirmed SuccessTool & computer use › General agents15.711.6−2.1
CritPt v2 (agentic, avg@5)Science › Physical sciences80.682.6+2.0
DualEntry Accounting AI BenchmarkProfessional › Finance, accounting & tax83.281.2−2.0
AA-Omniscience v1.1 IndexKnowledge & reasoning › General knowledge46.342.4−1.9
TxBench Oligo Discovery V1 (113 evaluations)Science › Life sciences54.352.5−1.8
ARC-AGI-2Knowledge & reasoning › Abstract reasoning92.594.2+1.7
TxBench Preclinical Pharmacology V1 Clean RerunsScience › Life sciences61.359.7−1.7
Epoch Capabilities Index (ECI)Composite indexes167169+1.4
MultiChallengeLanguage & writing › Conversation80.682.0+1.4
GradeBenchProfessional › Education51.953.2+1.3
LMArena Text Arena EloLanguage & writing › General assistance14901483−1.2
DAYJOB: Healthcare (October 2026 rescoring)Health & medicine24.025.2+1.2
Vibe Code Bench v1.1Software › Frontend & design87.888.9+1.2
Terminal-Bench-Science 0.1Science › Multidisciplinary49.550.5+1.0
Agents' Last Exam v1.1 ScoreProfessional › General knowledge work71.672.5+0.9
GDP.pdf (Artificial Analysis, All-pass)Professional › General knowledge work32.832.0−0.8
Terminal-Bench 4.0 (Artificial Analysis)Software › Software engineering53.554.0+0.5
FrontierCode 2.0 (Artificial Analysis)Software › Software engineering39.038.5−0.4
DAYJOB: TaxProfessional › Finance, accounting & tax3.64.0+0.4
LMArena Agent Arena Tool HallucinationTool & computer use › Tool calling0.71.2+0.2
ARC-AGI-1 Semi-PrivateKnowledge & reasoning › Abstract reasoning98.598.50.0
Dual-Arm Sim (task completion)Tool & computer use › Robotics0.00.00.0
Humanity's Last Exam DiamondKnowledge & reasoning › General knowledge58.858.80.0
OTIS Mock AIME 2024-2025Math100.0100.00.0
Data: Artificial Analysis, Vals AI, Epoch AI and 19 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…