Atlas

Compare Models

Gemini 4 Carbon vs GPT-5.6 Sol

Gemini 4 Carbon leads by 5.1 capability points (≈95% interval +2.5 to +7.7): a clear lead over GPT-5.6 Sol · P(higher) ≈ 100%

Wins 37 of 41 shared benchmarks (4 lost)

Capability vs. relative cost
Gemini 4 CarbonGPT-5.6 SolClosed 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 CarbonvsGPT-5.6 Solcapability vs. relative costData Nov 15, 2026Village LLM Atlasatlas.village.dev

Benchmark wins37 – 4 · 41 shared

Comparing matching setups only (17 benchmarks excluded) · suites counted once
← Gemini 4 CarbonGPT-5.6 Sol →

Gemini 4 CarbonvsGPT-5.6 Solwins by categoryData Nov 15, 2026Data: Artificial Analysis, Arena, benchmarks.bio and 13 othersVillage LLM Atlasatlas.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 Carbon.

ARC-AGI-3 (Semi-Private)General QA65.10.3−64.8
Terminal-Bench 4.0 (Artificial Analysis)Code60.614.6−46.0
TxBench-Antibody-Development V1 (60 evaluations)Science61.118.9−42.2
Terminal-Bench-Science 0.1Science62.422.4−40.0
Humanity's Last Exam DiamondGeneral QA69.130.4−38.7
FrontierSWE V2 Worst@5Code56.222.2−34.0
TxBench Antibody Discovery V1 (100 evaluations)Science65.037.3−27.7
ChartographyMultimodal72.545.0−27.5
Terminal-Bench 4.0Code63.137.9−25.3
Blueprint-Bench 2Multimodal0.60.3−25.0
AutomationBench 1.0.6 (Zapier)Agents53.028.8−24.2
LMArena Agent Arena Net Improvement (Steering Burden)Agents17.04.6−23.3
FrontierSWE V2 Best@5Code66.744.2−22.5
ExploitGym - release-reported percentage, cohort unspecifiedCybersecurity54.833.7−21.1
CursorBench 4.0Code60.441.7−18.7
TxBench Oligo Discovery V1 (113 evaluations)Science55.837.2−18.6
AutomationBench-AA 1.0.6Agents77.860.1−17.7
BuseyBench SVGMultimodal7.96.1−17.5
LMArena Agent Arena Praise vs ComplaintAgents41.06.3−17.3
AA-Briefcase v1.1 EloAgents16111369−13.2
Medical Long Context Reasoning (MLCR-AA, October 2026 methodology)Health & Medicine19.430.0+10.6
GDPval-AA v2.1Professional Work17331511−9.8
SEC-Bench Pro - release-reported cohortCybersecurity88.279.1−9.1
Vibe Code Bench v1.1Code89.180.5−8.6
Tax Agent BenchFinance & Accounting76.468.0−8.5
LMArena Agent Arena Bash ReliabilityAgents-5.211.4+8.3
LMArena Agent Arena Confirmed SuccessAgents18.42.3−8.0
Legal Research BenchLegal55.848.1−7.7
LMArena Text Arena EloChat & Writing15261485−7.2
Artificial Analysis Progress Index v1Indexes351331−6.8
LMArena Agent Arena Steering BurdenAgents19.15.5−6.8
MMMU ProMultimodal89.282.7−6.6
RiemannBenchMath80.074.4−5.6
CritPt v2 (agentic, avg@5)Science82.377.1−5.1
SWE-sweepCode9.64.7−4.9
TxBench Preclinical Pharmacology V1 Clean RerunsScience66.061.3−4.7
AA-LCR v1.1Search79.384.0+4.7
GDP.pdf (Artificial Analysis, All-pass)Professional Work29.627.8−1.8
ARC-AGI-2General QA94.292.5−1.7
ARC-AGI-1 Semi-PrivateGeneral QA98.597.5−1.0
LMArena Agent Arena Tool HallucinationAgents0.51.0+0.2

Gemini 4 CarbonvsGPT-5.6 Solnormalised score marginsData Nov 15, 2026Data: Artificial Analysis, Arena, benchmarks.bio and 13 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…