Atlas

Compare Models

Claude Sonnet 5.5 vs Gemini 4 Carbon

Gemini 4 Carbon leads by 1.3 capability points (≈95% interval −1.4 to +4.1): statistically tied · P(higher) ≈ 83%

Wins 49 of 60 shared benchmarks (10 lost, 1 tied)

Capability vs. relative cost
Claude Sonnet 5.5Gemini 4 CarbonClosed 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.

Claude Sonnet 5.5vsGemini 4 Carboncapability vs. relative costData Nov 15, 2026Village LLM Atlasatlas.village.dev

Benchmark wins10 – 49 · 60 shared, 1 tied

Comparing matching setups only (3 benchmarks excluded) · suites counted once
← Claude Sonnet 5.5Gemini 4 Carbon →

Claude Sonnet 5.5vsGemini 4 Carbonwins by categoryData Nov 15, 2026Data: Artificial Analysis, Arena, Mercor and 14 othersVillage LLM Atlasatlas.village.dev

Margins

Each benchmark scaled 0–100% (its declared range, else the field’s range); Δ is Gemini 4 Carbon minus Claude Sonnet 5.5.

ARC-AGI-3 (Semi-Private)General QA35.596.9+61.4
Terminal-Bench-Science 0.1Science16.262.4+46.2
Terminal-Bench 4.0 (Artificial Analysis)Code29.860.6+30.8
TxBench-Antibody-Development V1 (60 evaluations)Science36.761.1+24.4
Medical Long Context Reasoning (MLCR-AA, October 2026 methodology)Health & Medicine41.719.4−22.2
Blueprint-Bench 2Multimodal0.40.6+19.3
Agents' Last Exam v1.1 ScoreAgents60.477.4+17.0
Humanity's Last Exam DiamondGeneral QA53.369.1+15.8
BuseyBench SVGMultimodal6.57.9+13.3
CritPt v2 (agentic, avg@5)Science70.082.3+12.3
AA-Omniscience v1.1 IndexIndexes30.553.5+11.5
ChartographyMultimodal61.672.5+10.9
TxBench Antibody Discovery V1 (100 evaluations)Science54.765.0+10.3
ARC-AGI-2General QA83.994.2+10.3
Artificial Analysis Legal IndexIndexes52.262.0+9.8
LMArena Text Arena EloChat & Writing14721526+9.3
LMArena Agent Arena Bash ReliabilityAgents12.7-5.2−9.0
GDPval-AA v2.1Professional Work15511733+8.0
DAXBench v2Code86.293.6+7.4
HLAB-Vals-1.1Legal10.417.5+7.1
Artificial Analysis Finance & Accounting IndexIndexes55.862.6+6.9
Artificial Analysis Engineering IndexIndexes58.565.3+6.8
APEX-Agents 1.1Professional Work83.790.4+6.7
OpenWorldEngineer (Mean@8)Code39.032.8−6.2
APEX-Accounting 1.1Finance & Accounting64.570.7+6.2
AutomationBench-AA 1.0.6Agents71.877.8+6.0
Bug Hunt BenchCode48.954.6+5.7
Artificial Analysis Economics IndexIndexes54.159.6+5.5
Artificial Analysis Intelligence Index v5.0.1Indexes54.560.0+5.5
LMArena Agent Arena Praise vs ComplaintAgents30.541.0+5.3
FrontierSWE V2 Worst@5Code50.956.2+5.2
FrontierSWE V2 Best@5Code71.766.7−5.0
CursorBench 4.0Code55.560.4+4.9
VIALS-AA (pass^3)Multimodal59.664.0+4.3
TxBench Oligo Discovery V1 (113 evaluations)Science51.955.8+3.8
GDP.pdf (Artificial Analysis, All-pass)Professional Work25.829.6+3.8
Artificial Analysis Intelligence Index v5.0Indexes54.257.7+3.5
MMMU ProMultimodal85.889.2+3.4
Vibe Code Bench v1.1Code92.489.1−3.3
RiemannBench (November 2026 rescoring)Math76.880.0+3.2
Artificial Analysis Strategy & Ops IndexIndexes58.160.9+2.9
GDPxProfessional Work8.210.9+2.7
Artificial Analysis Healthcare & Medical IndexIndexes54.256.8+2.6
MKA (Artificial Analysis)Professional Work52.054.3+2.4
LMArena Agent Arena Steering BurdenAgents14.519.1+2.3
ARC-AGI-1 Semi-PrivateGeneral QA96.598.5+2.0
LMArena Agent Arena Net Improvement (Steering Burden)Agents16.017.0+1.9
VIALS-AA (average of 3)Multimodal65.467.3+1.9
τ³-Banking (Mercor run, bm25_grep retrieval)Agents54.055.7+1.7
TxBench Preclinical Pharmacology V1 Clean RerunsScience64.366.0+1.7
AA-Briefcase v1.1 EloAgents16341611−1.3
LMArena Code Arena WebDevCode17731768−0.6
Excel Modeling Benchmark v1.1Finance & Accounting75.976.5+0.6
LMArena Agent Arena Confirmed SuccessAgents19.518.4−0.6
FrontierCode 2.0 - Pass RateCode46.446.9+0.6
FrontierCode 2.0 - ScoreCode42.242.6+0.4
AA-LCR v1.1Search79.779.3−0.3
LMArena Agent Arena Tool HallucinationAgents0.90.5−0.2
Artificial Analysis Progress Index v1Indexes351351+0.1
VIALS-AA (pass@3)Multimodal70.870.80.0

Claude Sonnet 5.5vsGemini 4 Carbonnormalised score marginsData Nov 15, 2026Data: Artificial Analysis, Arena, Mercor 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…