Atlas

Compare Models

Claude Haiku 5.5 vs Gemini 4 Carbon

Gemini 4 Carbon leads by 6.6 capability points (≈95% interval +4.1 to +9.2): a clear lead over Claude Haiku 5.5 · P(higher) ≈ 100%

Wins 57 of 61 shared benchmarks (4 lost)

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

Benchmark wins4 – 57 · 61 shared

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

Claude Haiku 5.5vsGemini 4 Carbonwins by categoryData Nov 15, 2026Data: Artificial Analysis, Vals AI, Arena and 11 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 Haiku 5.5.

ARC-AGI-3 (Semi-Private)General QA2.899.1+96.3
Terminal-Bench-Science 0.1Science7.162.4+55.2
Terminal-Bench 4.0 (Artificial Analysis)Code15.260.6+45.5
AutomationBench-AA 1.0.6Agents36.077.8+41.9
AutomationBench 1.0.6 (Zapier)Agents12.053.0+40.9
Humanity's Last Exam DiamondGeneral QA29.669.1+39.5
Bug Hunt BenchCode20.554.6+34.1
BuseyBench SVGMultimodal4.67.9+33.0
ARC-AGI-2General QA61.394.2+32.9
LMArena Agent Arena Net Improvement (Steering Burden)Agents0.517.0+31.2
Code Migration v1.1 (Vals Index v2.2 subset)Code50.280.1+29.9
Terminal-Bench 4.0Code36.463.1+26.8
Blueprint-Bench 2Multimodal0.30.6+25.6
Artificial Analysis Intelligence Index v5.0.1Indexes36.560.0+23.5
Artificial Analysis Legal IndexIndexes38.762.0+23.3
LMArena Code Arena WebDevCode15941768+22.8
Artificial Analysis Engineering IndexIndexes42.565.3+22.8
LMArena Agent Arena Praise vs ComplaintAgents-3.541.0+22.3
APEX-Agents 1.1Professional Work68.390.4+22.1
Artificial Analysis Strategy & Ops IndexIndexes39.160.9+21.8
FrontierSWE V2 Worst@5Code34.556.2+21.7
Artificial Analysis Intelligence Index v5.0Indexes36.257.7+21.5
AA-Omniscience v1.1 IndexIndexes12.353.5+20.6
CritPt v2 (agentic, avg@5)Science62.682.3+19.7
Artificial Analysis Healthcare & Medical IndexIndexes37.556.8+19.3
Artificial Analysis Finance & Accounting IndexIndexes43.662.6+19.1
Harvey LAB (Mercor run, undisclosed ~213-task cohort)Legal4.923.0+18.1
LMArena Text Arena EloChat & Writing14221526+18.1
Artificial Analysis Economics IndexIndexes42.759.6+16.9
Medical Long Context Reasoning (MLCR-AA, October 2026 methodology)Health & Medicine35.019.4−15.6
FrontierCode 2.0 - Pass RateCode31.946.9+15.1
VIALS-AA (pass^3)Multimodal49.164.0+14.9
Tax Agent BenchFinance & Accounting62.276.4+14.2
FrontierSWE V2 Best@5Code52.866.7+13.9
FrontierCode 2.0 - ScoreCode28.942.6+13.7
GDPval-AA v2.1Professional Work14241733+13.6
MKA (Artificial Analysis)Professional Work41.054.3+13.4
HLAB-Vals-1.1Legal4.217.5+13.3
τ³-Banking (Mercor run, bm25_grep retrieval)Agents43.055.7+12.7
Legal Research BenchLegal43.855.8+12.0
CursorBench 4.0Code48.460.4+12.0
DAXBench v2Code81.993.6+11.7
Excel Modeling Benchmark v1.1Finance & Accounting65.576.5+11.0
MMMU ProMultimodal79.189.2+10.1
Vals Finance IndexIndexes57.467.0+9.6
ARC-AGI-1 Semi-PrivateGeneral QA89.098.5+9.5
AA-Briefcase v1.1 EloAgents14411611+9.3
GDP.pdf (Artificial Analysis, All-pass)Professional Work20.829.6+8.8
Artificial Analysis Progress Index v1Indexes326351+8.7
APEX-Accounting 1.1Finance & Accounting62.270.7+8.5
LMArena Agent Arena Confirmed SuccessAgents1.618.4+8.4
LMArena Agent Arena Steering BurdenAgents2.419.1+8.4
VIALS-AA (average of 3)Multimodal60.567.3+6.8
LMArena Agent Arena Bash ReliabilityAgents7.4-5.2−6.3
GDPxProfessional Work4.910.9+5.9
Vals Legal IndexIndexes70.375.9+5.6
Finance Agent v2.1Finance & Accounting65.467.2+1.8
Vibe Code Bench v1.1Code90.489.1−1.4
LMArena Agent Arena Tool HallucinationAgents-2.10.5+1.3
AA-LCR v1.1Search78.379.3+1.0
VIALS-AA (pass@3)Multimodal71.470.8−0.6

Claude Haiku 5.5vsGemini 4 Carbonnormalised score marginsData Nov 15, 2026Data: Artificial Analysis, Vals AI, Arena and 11 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…