Atlas

Compare Models

Gemini 4 Carbon vs Grok 4.7

Gemini 4 Carbon leads by 7.3 capability points (≈95% interval +4.8 to +9.9): a clear lead over Grok 4.7 · P(higher) ≈ 100%

Wins 66 of 72 shared benchmarks (6 lost)

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

Each line runs through one model’s priced reasoning settings, cheapest to most expensive; grey points are other models’ priced settings in the same range. Results whose reasoning effort the source didn’t report are left out wherever a model has named effort levels. Capability and relative cost are the published estimates (relative cost is cost per task as a multiple of GPT-5 (high)); hover a setting, or use the arrow keys, to set it beside the other model’s same setting.

Gemini 4 CarbonvsGrok 4.7capability vs. relative costData Nov 15, 2026Village LLM Atlasatlas.village.dev

Benchmark wins66 – 6 · 72 shared

Comparing matching setups only (6 benchmarks excluded) · suites counted once
← Gemini 4 CarbonGrok 4.7 →

Gemini 4 CarbonvsGrok 4.7wins by categoryData Nov 15, 2026Data: Artificial Analysis, Vals AI, Arena and 14 othersVillage LLM Atlasatlas.village.dev

Margins

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

ARC-AGI-3 (Semi-Private)General QA99.18.4−90.7
ChartographyMultimodal72.514.7−57.8
Terminal-Bench-Science 0.1Science61.94.8−57.1
BrokenArXiv 09/2026 (Native Harnesses)Math82.427.7−54.7
BrokenArXiv 08/2026 (Native Harnesses)Math88.135.1−53.0
ArXivMath 09/2026 (Native Harnesses)Math96.245.3−50.9
ArXivMath 08/2026 (Native Harnesses)Math93.046.5−46.5
Terminal-Bench 4.0 (Artificial Analysis)Code62.616.2−46.5
Humanity's Last Exam DiamondGeneral QA69.123.6−45.5
FrontierSWE V2 Worst@5Code56.215.9−40.3
Terminal-Bench 4.0Code63.128.8−34.3
BuseyBench SVGMultimodal7.94.6−32.9
ARC-AGI-2General QA94.261.4−32.8
Code Migration v1.1 (Vals Index v2.2 subset)Code80.151.8−28.2
LMArena Agent Arena Net Improvement (Steering Burden)Agents17.02.6−27.2
Bug Hunt BenchCode54.627.6−27.0
Blueprint-Bench 2Multimodal0.60.3−26.1
VIALS-AA (pass^3)Multimodal64.037.9−26.1
OpenWorldEngineer (Mean@8)Code32.87.0−25.7
FrontierSWE V2 Best@5Code66.741.1−25.6
Artificial Analysis Intelligence Index v5.0.1Indexes60.035.5−24.5
Artificial Analysis Engineering IndexIndexes65.341.8−23.5
τ³-Banking (Mercor run, bm25_grep retrieval)Agents55.732.6−23.1
RiemannBench (November 2026 rescoring)Math80.057.6−22.4
Artificial Analysis Intelligence Index v5.0Indexes57.735.3−22.4
APEX-Agents 1.1Professional Work90.468.7−21.7
LMArena Agent Arena Praise vs ComplaintAgents41.03.1−19.0
CritPt v2 (agentic, avg@5)Science79.762.3−17.4
LMArena Code Arena WebDevCode17681641−16.7
APEX-Accounting 1.1Finance & Accounting70.755.3−15.4
FrontierBioMLScience39.024.0−15.0
VIALS-AA (average of 3)Multimodal67.352.4−14.9
Artificial Analysis Legal IndexIndexes62.047.2−14.8
LMArena Text Arena EloChat & Writing15261442−14.7
FrontierCode 2.0 - Pass RateCode46.932.3−14.6
MKA (Artificial Analysis)Professional Work54.340.1−14.2
CursorBench 4.0Code60.446.3−14.1
Artificial Analysis Economics IndexIndexes59.645.7−13.9
Artificial Analysis Coding Agent Index v2.1Indexes55.541.6−13.8
FrontierCode 2.0 - ScoreCode42.628.9−13.7
Artificial Analysis Coding Agent Index v2.0Indexes59.446.6−12.8
AutomationBench-AA 1.0.6Agents77.865.6−12.3
AA-Omniscience v1.1 IndexIndexes53.531.2−11.1
MMMU ProMultimodal89.278.2−11.1
Artificial Analysis Finance & Accounting IndexIndexes62.651.7−10.9
LMArena Agent Arena Steering BurdenAgents19.1-2.3−10.7
Artificial Analysis Healthcare & Medical IndexIndexes56.846.6−10.2
Artificial Analysis Progress Index v1Indexes351323−9.9
Legal Research BenchLegal55.847.1−8.7
ARC-AGI-1 Semi-PrivateGeneral QA98.590.2−8.3
Vals Finance IndexIndexes67.058.8−8.2
Vals RSI Index v1.1 (four tasks)Agents27.720.2−7.5
Tax Agent BenchFinance & Accounting76.469.0−7.4
LMArena Agent Arena Bash ReliabilityAgents-5.29.3+7.3
GDPxProfessional Work10.93.9−7.0
LMArena Agent Arena Confirmed SuccessAgents18.45.1−6.6
Finance Agent v2.1Finance & Accounting67.260.6−6.6
GDP.pdf (Artificial Analysis, All-pass)Professional Work29.623.2−6.4
Artificial Analysis Strategy & Ops IndexIndexes60.954.8−6.1
Excel Modeling Benchmark v1.1Finance & Accounting76.571.5−5.0
VIALS-AA (pass@3)Multimodal70.865.8−5.0
Harvey LAB (Mercor run, undisclosed ~213-task cohort)Legal23.019.5−3.5
Vals Legal IndexIndexes75.972.7−3.3
CWE-bench v1 Programmatic Pass@1Cybersecurity71.068.0−3.0
Vibe Code Bench v1.1Code89.186.2−2.9
Medical Long Context Reasoning (MLCR-AA, October 2026 methodology)Health & Medicine19.421.1+1.7
DAXBench v2Code93.692.6−1.1
AA-Briefcase v1.1 EloAgents16111629+1.0
HLAB-Vals-1.1Legal17.518.3+0.8
AA-LCR v1.1Search79.380.0+0.7
GDPval-AA v2.1Professional Work17331718−0.6
LMArena Agent Arena Tool HallucinationAgents0.51.0+0.2

Gemini 4 CarbonvsGrok 4.7normalised score marginsData Nov 15, 2026Data: Artificial Analysis, Vals AI, Arena and 14 othersVillage LLM Atlasatlas.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…