Atlas

Compare Models

Gemini 4 Carbon vs GPT-6.2 Astra

GPT-6.2 Astra leads by 0.8 capability points (≈95% interval −2.0 to +3.7): statistically tied · P(higher) ≈ 72%

Wins 44 of 84 shared benchmarks (36 lost, 4 tied)

Capability vs. relative cost
Gemini 4 CarbonGPT-6.2 AstraClosed weightsOpen weights

Benchmark wins36 – 44 · 84 shared, 4 tied

Comparing matching setups only (41 benchmarks excluded) · suites counted once
← Gemini 4 CarbonGPT-6.2 Astra →
Data: Artificial Analysis, Surge AI, Epoch AI and 18 othersatlas.village.dev

Margins

Each benchmark scaled 0–100% (its declared range, else the field’s range); Δ is GPT-6.2 Astra minus Gemini 4 Carbon.

KradleBenchGames22.078.3+56.3
EnterpriseBench: CoreCraftTool & computer use › Customer service & operations80.543.6−36.9
DAYJOB: Law (November 2026 release)Professional › Legal13.841.4+27.6
DeepsecBench — Revalidated V1 (233 issues)Software › Cybersecurity35.261.2+26.0
Mystery Game PuzzlesGames73.095.0+22.0
DAYJOB: Finance (October 2026 rescoring)Professional › Finance, accounting & tax23.845.8+22.0
Medical Long Context Reasoning (MLCR-AA, October 2026 methodology)Health & medicine19.439.4+20.0
Tax Agent BenchProfessional › Finance, accounting & tax76.457.9−18.6
DAYJOB: AccountingProfessional › Finance, accounting & tax25.242.0+16.8
SimpleBenchKnowledge & reasoning › General knowledge92.677.3−15.3
Legal Research BenchProfessional › Legal55.842.8−13.0
MultiChallengeLanguage & writing › Conversation85.973.2−12.7
Terminal-Bench-Science 0.1Science › Multidisciplinary62.474.8+12.4
Furniture Assembly BenchmarkPerception & media › Spatial & visual reasoning77.589.2+11.7
Dual-Arm Sim (task completion)Tool & computer use › Robotics13.82.5−11.3
SWE Atlas v1.1 - OverallSoftware › Software engineering71.560.5−11.0
Professional Reasoning Benchmark - LegalProfessional › Legal53.443.1−10.3
Vals Finance IndexProfessional › Finance, accounting & tax67.057.3−9.7
GDP.xlsxProfessional › General knowledge work38.929.7−9.2
Humanity's Sixth SensePerception & media › Spatial & visual reasoning71.163.2−7.9
sudo L7 (October 2026 rescoring)Software › Software engineering30.638.3+7.7
FrontierMath ErdősMath2.910.3+7.4
GDP.pdf (November 2026 rescoring)Professional › General knowledge work56.449.2−7.2
LMArena Agent Arena Bash ReliabilityTool & computer use › General agents-5.77.7+6.7
FrontierCode 2.0 (Artificial Analysis)Software › Software engineering40.647.1+6.6
Finance Agent v2.1Professional › Finance, accounting & tax67.260.7−6.4
Vals Legal IndexProfessional › Legal75.969.7−6.2
DrugDiscoveryBenchScience › Life sciences71.565.5−6.0
Chartography (November 2026 rescoring)Perception & media › Documents & charts85.590.7+5.2
FrontierMath Tier 4 (v2) - PrivateMath95.1100.0+4.9
EBR-bench v4 (card ban, single agent)Games87.692.4+4.8
DAYJOB: TaxProfessional › Finance, accounting & tax4.89.2+4.4
HANDBOOK.md Agents v1.1Tool & computer use › Customer service & operations49.845.8−4.0
RiemannBench (November 2026 rescoring)Math80.076.0−4.0
BigFinanceBench private holdout (final-answer accuracy)Professional › Finance, accounting & tax50.154.0+3.9
VIALS-AA (pass@3)Science › Life sciences70.874.5+3.7
Code Migration v1.1 (Vals Index v2.2 subset)Software › Software engineering80.183.6+3.5
Artificial Analysis Legal IndexProfessional › Legal62.058.7−3.2
Humanity's Last Exam DiamondKnowledge & reasoning › General knowledge69.172.2+3.1
VIALS-AA (average of 3)Science › Life sciences67.370.2+2.9
Blueprint-Bench 2Perception & media › Spatial & visual reasoning0.60.6+2.8
Epoch Capabilities Index (ECI)Composite indexes170173+2.5
LMArena Agent Arena Confirmed SuccessTool & computer use › General agents17.212.4−2.4
Artificial Analysis Strategy & Ops IndexProfessional › General knowledge work60.963.3+2.3
AA-Briefcase v1.1 EloProfessional › General knowledge work16111570−2.2
Artificial Analysis Intelligence Index v5.0.1Composite indexes60.062.1+2.1
FrontierMath Tiers 1-3 (v2) - PrivateMath90.592.6+2.1
Terminal-Bench 4.0 (Artificial Analysis)Software › Software engineering60.658.6−2.0
TxBench Preclinical Pharmacology V1 Clean RerunsScience › Life sciences66.068.0+2.0
LMArena Agent Arena Steering BurdenTool & computer use › General agents17.914.0−2.0
VIALS-AA (pass^3)Science › Life sciences64.065.8+1.9
AA-Omniscience v1.1 IndexKnowledge & reasoning › General knowledge53.550.0−1.7
LMArena Agent Arena Net Improvement (Steering Burden)Tool & computer use › General agents16.115.2−1.7
Artificial Analysis Finance & Accounting IndexProfessional › Finance, accounting & tax62.660.9−1.7
MKA (Artificial Analysis)Professional › General knowledge work54.352.7−1.7
ARC-AGI-2Knowledge & reasoning › Abstract reasoning94.295.8+1.7
GDPval v2 (OpenAI internal)Professional › General knowledge work48.850.4+1.6
LiveBench 2.0 (2026-11-06 release)Composite indexes83.384.9+1.5
Artificial Analysis Economics IndexComposite indexes59.661.1+1.5
SEC-Bench Pro - release-reported cohortSoftware › Cybersecurity88.286.8−1.5
CritPt v2 (agentic, avg@5)Science › Physical sciences82.383.7+1.4
Artificial Analysis Engineering IndexScience › Engineering65.366.6+1.3
ComplexConstraints (November 2026 rescoring)Language & writing › Instruction following70.969.6−1.3
ARC-AGI-3 (Semi-Private)Games99.197.8−1.3
HLAB-Vals-1.1Professional › Legal17.516.3−1.2
LMArena Agent Arena Praise vs ComplaintTool & computer use › General agents39.437.0−1.2
Ramp SWE-Bench v2 (Bash harness)Software › Software engineering74.075.0+1.0
Agents' Last Exam v1.1 ScoreProfessional › General knowledge work77.478.3+0.9
Vibe Code Bench v1.1Software › Frontend & design89.189.9+0.8
DAYJOB: Healthcare (October 2026 rescoring)Health & medicine35.234.4−0.8
SimpleQA Verified (Epoch anti-abstention)Knowledge & reasoning › General knowledge77.778.5+0.8
AutomationBench 1.0.6 (Zapier)Tool & computer use › Customer service & operations53.052.4−0.6
AutomationBench-AA 1.0.6Tool & computer use › Customer service & operations77.877.2−0.6
Excel Modeling Benchmark v1.1Professional › Finance, accounting & tax76.576.0−0.4
LLM Debate BenchmarkLanguage & writing › Language & wordplay16761673−0.4
GPQA DiamondScience › Multidisciplinary96.596.8+0.3
MMMU ProKnowledge & reasoning › General knowledge89.289.5+0.2
Artificial Analysis Healthcare & Medical IndexHealth & medicine56.857.0+0.2
GDP.pdf (Artificial Analysis, All-pass)Professional › General knowledge work29.629.8+0.2
LMArena Agent Arena Tool HallucinationTool & computer use › Tool calling0.81.2+0.2
ARC-AGI-1 Semi-PrivateKnowledge & reasoning › Abstract reasoning98.598.50.0
Chess Puzzles (task version 2.0)Games78.078.00.0
OTIS Mock AIME 2024-2025Math100.0100.00.0
Terminal-Bench 4.0Software › Software engineering63.163.10.0
Data: Artificial Analysis, Surge AI, Epoch AI and 18 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…