Atlas

Compare Models

Gemini 4 Hydrogen vs Grok 4.7

Gemini 4 Hydrogen leads by 5.1 capability points (≈95% interval +2.6 to +7.7): a clear lead over Grok 4.7 · P(higher) ≈ 100%

Wins 40 of 46 shared benchmarks (5 lost, 1 tied)

Capability vs. relative cost
Gemini 4 HydrogenGrok 4.7Closed weightsOpen weights

Benchmark wins40 – 5 · 46 shared, 1 tied

Comparing matching setups only (3 benchmarks excluded) · suites counted once
← Gemini 4 HydrogenGrok 4.7 →
Data: Artificial Analysis, Vals AI, Arena and 10 othersatlas.village.dev

Margins

Each benchmark scaled 0–100% (its declared range, else the field’s range); Δ is Grok 4.7 minus Gemini 4 Hydrogen.

ChartographyMultimodal66.614.7−51.9
Terminal-Bench-Science 0.1Science49.511.4−38.1
Humanity's Last Exam DiamondGeneral QA58.823.6−35.2
Code Migration v1.1 (Vals Index v2.2 subset)Code82.751.8−30.9
Terminal-Bench 4.0 (Artificial Analysis)Code53.525.8−27.8
Terminal-Bench 4.0Code54.028.8−25.3
LMArena Agent Arena Net Improvement (Steering Burden)Agents14.91.8−25.0
FrontierSWE V2Code53.129.5−23.6
BuseyBench SVGMultimodal6.94.6−22.9
CritPt v2 (agentic, avg@5)Science78.055.7−22.3
Artificial Analysis Engineering IndexIndexes60.141.8−18.3
GradeBenchProfessional Work51.935.3−16.6
LMArena Agent Arena Praise vs ComplaintAgents33.82.2−15.8
LMArena Agent Arena Steering BurdenAgents18.4-3.6−11.0
FrontierCode 2.0 - Pass RateCode43.232.3−10.9
Artificial Analysis Coding Agent Index v2.1Indexes52.341.6−10.7
FrontierCode 2.0 - ScoreCode38.928.9−10.0
Vals Index v2.2Indexes68.558.6−9.9
GDP.pdf (Artificial Analysis, All-pass)Professional Work32.823.2−9.6
Artificial Analysis Legal IndexIndexes56.747.2−9.5
Tax Agent BenchFinance & Accounting78.469.0−9.4
Artificial Analysis Economics IndexIndexes55.045.7−9.4
AutomationBench-AA 1.0.6Agents74.465.6−8.9
LMArena Code Arena WebDevCode17041640−8.5
LMArena Text Arena EloChat & Writing14901443−8.2
CursorBench 4.0Code54.346.3−8.0
AA-Omniscience v1.1 IndexIndexes46.331.2−7.5
Vals Finance IndexIndexes66.158.8−7.3
HLAB-Vals-1.1Legal11.318.3+7.1
Artificial Analysis Healthcare & Medical IndexIndexes53.346.6−6.7
Legal Research BenchLegal53.847.1−6.7
LMArena Agent Arena Bash ReliabilityAgents-4.68.4+6.5
Artificial Analysis Finance & Accounting IndexIndexes58.151.7−6.4
GBENCH One-shot Coding GScoreGames66.860.5−6.3
LMArena Agent Arena Confirmed SuccessAgents16.84.2−6.3
Artificial Analysis Intelligence Index v5.0.1Indexes41.235.5−5.7
DualEntry Accounting AI BenchmarkFinance & Accounting83.288.1+4.9
Finance Agent v2.1Finance & Accounting65.160.6−4.5
Excel Modeling Benchmark v1.1Finance & Accounting75.871.5−4.3
AA-Briefcase v1.1 EloAgents15681636+3.7
Bug Hunt BenchCode31.127.6−3.5
Artificial Analysis Strategy & Ops IndexIndexes57.754.8−2.9
Vibe Code Bench v1.1Code87.886.2−1.6
Vals Legal IndexIndexes73.672.7−1.0
LMArena Agent Arena Tool HallucinationAgents0.51.0+0.3
Medical Long Context Reasoning (MLCR-AA, October 2026 methodology)Health & Medicine21.121.10.0
Data: Artificial Analysis, Vals AI, Arena and 10 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…