Atlas

Compare Models

Gemini 4 Hydrogen vs GPT-6 Sol

Gemini 4 Hydrogen leads by 2.1 capability points (≈95% interval −0.3 to +4.6): statistically tied · P(higher) ≈ 95%

Wins 37 of 52 shared benchmarks (14 lost, 1 tied)

Capability vs. relative cost
Gemini 4 HydrogenGPT-6 SolClosed weightsOpen weights

Benchmark wins37 – 14 · 52 shared, 1 tied

Comparing matching setups only (7 benchmarks excluded) · suites counted once
← Gemini 4 HydrogenGPT-6 Sol →
Data: Artificial Analysis, Epoch AI, Arena and 15 othersatlas.village.dev

Margins

Each benchmark scaled 0–100% (its declared range, else the field’s range); Δ is GPT-6 Sol minus Gemini 4 Hydrogen.

Humanity's Sixth SensePerception & media › Spatial & visual reasoning63.431.2−32.2
Terminal-Bench-Science 0.1Science › Multidisciplinary49.517.6−31.9
Tax Agent BenchProfessional › Finance, accounting & tax78.453.0−25.4
Legal Research BenchProfessional › Legal53.828.8−25.0
GDP.xlsxProfessional › General knowledge work40.917.1−23.8
Terminal-Bench 4.0 (Artificial Analysis)Software › Software engineering53.530.3−23.2
DAYJOB: Finance (October 2026 rescoring)Professional › Finance, accounting & tax38.519.5−19.0
SimpleBenchKnowledge & reasoning › General knowledge91.773.1−18.6
AutomationBench 1.0.6 (Zapier)Tool & computer use › Customer service & operations51.133.2−17.9
EBR-bench v4 (card ban, single agent)Games70.553.3−17.1
DAYJOB: Healthcare (October 2026 rescoring)Health & medicine24.07.2−16.8
Professional Reasoning Benchmark - LegalProfessional › Legal53.138.6−14.5
TxBench Antibody Discovery V1 (100 evaluations)Science › Life sciences57.743.3−14.3
TaxCalcBench TY25Professional › Finance, accounting & tax44.030.0−14.0
LMArena Agent Arena Net Improvement (Steering Burden)Tool & computer use › General agents13.56.5−13.8
Humanity's Last Exam DiamondKnowledge & reasoning › General knowledge58.845.2−13.6
ChartographyPerception & media › Documents & charts66.653.6−13.0
AutomationBench-AA 1.0.6Tool & computer use › Customer service & operations74.461.7−12.7
AA-Omniscience v1.1 IndexKnowledge & reasoning › General knowledge46.321.4−12.4
AA-Briefcase v1.1 EloProfessional › General knowledge work15691359−11.5
TxBench Oligo Discovery V1 (113 evaluations)Science › Life sciences54.343.1−11.2
SimpleQA Verified (Epoch anti-abstention)Knowledge & reasoning › General knowledge70.960.7−10.2
Gray Swan IPI Attack Success (k=15)Safety & alignment0.010.1−10.1
LMArena Agent Arena Praise vs ComplaintTool & computer use › General agents29.39.8−9.8
Mystery Game PuzzlesGames65.056.0−9.0
GDP.pdf (Artificial Analysis, All-pass)Professional › General knowledge work32.825.2−7.6
LMArena Agent Arena Bash ReliabilityTool & computer use › General agents-4.510.0+7.3
DualEntry Accounting AI BenchmarkProfessional › Finance, accounting & tax83.276.2−7.0
BuseyBench SVGPerception & media › Generation6.97.5+6.7
Medical Long Context Reasoning (MLCR-AA, October 2026 methodology)Health & medicine27.233.9+6.7
LMArena Agent Arena Confirmed SuccessTool & computer use › General agents15.73.9−5.9
LMArena Text Arena EloLanguage & writing › General assistance14901458−5.6
Terminal-Bench 4.0Software › Software engineering54.049.4−4.6
LMArena Agent Arena Steering BurdenTool & computer use › General agents17.49.3−4.1
Chess Puzzles (task version 2.0)Games60.056.0−4.0
FORTRESS Average Risk Score (ARS)Safety & alignment17.513.5+4.0
FrontierCode 2.0 (Artificial Analysis)Software › Software engineering39.035.6−3.4
ARC-AGI-1 Semi-PrivateKnowledge & reasoning › Abstract reasoning98.595.5−3.0
Artificial Analysis Intelligence Index v5.0.1Composite indexes41.244.2+2.9
ARC-AGI-2Knowledge & reasoning › Abstract reasoning92.589.6−2.9
Epoch Capabilities Index (ECI)Composite indexes167164−2.4
BigFinanceBench private holdout (final-answer accuracy)Professional › Finance, accounting & tax49.447.1−2.3
FrontierMath Tier 4 (v2) - PrivateMath87.890.0+2.2
GBENCH One-shot Coding GScoreGames66.868.8+2.0
FrontierMath Tiers 1-3 (v2) - PrivateMath88.189.8+1.8
TxBench Preclinical Pharmacology V1 Clean RerunsScience › Life sciences61.363.0+1.7
SpireClimbGames7.79.2+1.5
Furniture Assembly BenchmarkPerception & media › Spatial & visual reasoning57.558.3+0.8
CritPt v2 (agentic, avg@5)Science › Physical sciences80.680.9+0.3
LMArena Agent Arena Tool HallucinationTool & computer use › Tool calling0.71.2+0.2
Vibe Code Bench v1.1Software › Frontend & design87.887.80.0
OTIS Mock AIME 2024-2025Math100.0100.00.0
Data: Artificial Analysis, 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…