Atlas

Compare Models

Gemini 4 Hydrogen vs GPT-5.6 Sol

Gemini 4 Hydrogen leads by 3.1 capability points (≈95% interval +0.7 to +5.5): a clear lead over GPT-5.6 Sol · P(higher) ≈ 99%

Wins 43 of 50 shared benchmarks (4 lost, 3 tied)

Capability vs. relative cost
Gemini 4 HydrogenGPT-5.6 SolClosed weightsOpen weights

Benchmark wins43 – 4 · 50 shared, 3 tied

Comparing matching setups only (13 benchmarks excluded) · suites counted once
← Gemini 4 HydrogenGPT-5.6 Sol →
Data: Epoch AI, Artificial Analysis, Arena and 15 othersatlas.village.dev

Margins

Each benchmark scaled 0–100% (its declared range, else the field’s range); Δ is GPT-5.6 Sol minus Gemini 4 Hydrogen.

TxBench-Antibody-Development V1 (60 evaluations)Science › Life sciences60.618.9−41.7
Humanity's Sixth SensePerception & media › Spatial & visual reasoning63.430.0−33.4
Terminal-Bench-Science 0.1Science › Multidisciplinary51.422.4−29.0
Terminal-Bench 4.0 (Artificial Analysis)Software › Software engineering53.524.7−28.8
Humanity's Last Exam DiamondKnowledge & reasoning › General knowledge58.830.4−28.4
Gray Swan IPI Attack Success (k=15)Safety & alignment0.027.0−27.0
DAYJOB: Finance (October 2026 rescoring)Professional › Finance, accounting & tax38.511.8−26.7
EBR-bench v4 (card ban, single agent)Games70.544.8−25.7
HiL-Bench (Human-in-Loop Benchmark)Tool & computer use › General agents55.332.3−23.0
AutomationBench 1.0.6 (Zapier)Tool & computer use › Customer service & operations51.128.8−22.4
ChartographyPerception & media › Documents & charts66.645.0−21.6
FrontierSWE V2Software › Software engineering53.132.2−20.9
CursorBench 4.1Software › Software engineering61.440.9−20.5
TxBench Antibody Discovery V1 (100 evaluations)Science › Life sciences57.737.3−20.3
DAYJOB: Healthcare (October 2026 rescoring)Health & medicine24.04.4−19.6
LMArena Agent Arena Net Improvement (Steering Burden)Tool & computer use › General agents13.53.7−19.2
TxBench Oligo Discovery V1 (113 evaluations)Science › Life sciences54.337.2−17.1
Terminal-Bench 4.0Software › Software engineering54.037.9−16.2
AutomationBench-AA 1.0.6Tool & computer use › Customer service & operations74.460.1−14.4
LMArena Agent Arena Praise vs ComplaintTool & computer use › General agents29.33.8−12.7
DrugDiscoveryBenchScience › Life sciences69.156.5−12.6
CursorBench 4.0Software › Software engineering54.341.7−12.6
TaxCalcBench TY25Professional › Finance, accounting & tax44.032.0−12.0
Tax Agent BenchProfessional › Finance, accounting & tax78.468.0−10.5
LMArena Agent Arena Bash ReliabilityTool & computer use › General agents-4.510.5+7.5
BuseyBench SVGPerception & media › Generation6.96.1−7.5
AA-Briefcase v1.1 EloProfessional › General knowledge work15691436−7.3
Vibe Code Bench v1.1Software › Frontend & design87.880.5−7.3
Mystery Game PuzzlesGames65.058.0−7.0
GPQA DiamondScience › Multidisciplinary96.889.9−6.9
LMArena Agent Arena Steering BurdenTool & computer use › General agents17.43.8−6.8
LMArena Agent Arena Confirmed SuccessTool & computer use › General agents15.72.5−6.6
Legal Research BenchProfessional › Legal53.848.1−5.8
BigFinanceBench private holdout (final-answer accuracy)Professional › Finance, accounting & tax49.444.2−5.2
GDP.pdf (Artificial Analysis, All-pass)Professional › General knowledge work32.827.8−5.0
FrontierMath Tier 4 (v2) - PrivateMath87.882.9−4.9
CritPt v2 (agentic, avg@5)Science › Physical sciences80.677.1−3.4
Epoch Capabilities Index (ECI)Composite indexes167163−3.0
Medical Long Context Reasoning (MLCR-AA, October 2026 methodology)Health & medicine27.230.0+2.8
FORTRESS Average Risk Score (ARS)Safety & alignment17.520.1−2.7
FrontierMath Tiers 1-3 (v2) - PrivateMath88.189.1+1.1
ARC-AGI-1 Semi-PrivateKnowledge & reasoning › Abstract reasoning98.597.5−1.0
Chess Puzzles (task version 2.0)Games60.059.0−1.0
Furniture Assembly BenchmarkPerception & media › Spatial & visual reasoning57.556.7−0.8
LMArena Text Arena EloLanguage & writing › General assistance14901485−0.8
GBENCH One-shot Coding GScoreGames66.866.4−0.4
LMArena Agent Arena Tool HallucinationTool & computer use › Tool calling0.71.2+0.2
ARC-AGI-2Knowledge & reasoning › Abstract reasoning92.592.50.0
OTIS Mock AIME 2024-2025Math100.0100.00.0
TxBench Preclinical Pharmacology V1 Clean RerunsScience › Life sciences61.361.30.0
Data: Epoch AI, Artificial Analysis, 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…