Atlas

Compare Models

Gemini 4 Argon vs GPT-6.2 Astra

GPT-6.2 Astra leads by 3.2 capability points (≈95% interval +0.4 to +6.1): a clear lead over Gemini 4 Argon · P(higher) ≈ 99%

Wins 38 of 52 shared benchmarks (12 lost, 2 tied)

Capability vs. relative cost
Gemini 4 ArgonGPT-6.2 AstraClosed 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 ArgonvsGPT-6.2 Astracapability vs. relative costData Nov 17, 2026Village LLM Atlasatlas.village.dev

Benchmark wins12 – 38 · 52 shared, 2 tied

Comparing matching setups only (7 benchmarks excluded) · suites counted once
← Gemini 4 ArgonGPT-6.2 Astra →

Gemini 4 ArgonvsGPT-6.2 Astrawins by categoryData Nov 17, 2026Data: Artificial Analysis, Vals AI, Epoch AI and 5 othersVillage LLM Atlasatlas.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 Argon.

AA-PersonalityBench Slop indexChat & Writing67.61.0−67.3
AA-PersonalityBench Informativeness indexChat & Writing82.922.0−61.5
AA-PersonalityBench Organization indexChat & Writing56.216.6−40.0
Mystery Game PuzzlesGames64.095.0+31.0
Humanity's Last Exam DiamondGeneral QA42.572.2+29.7
EBR-bench v4 (card ban, single agent)Games64.892.4+27.6
Terminal-Bench Science 0.1 (Vals)Science44.371.4+27.1
Terminal-Bench-Science 0.1Science48.674.8+26.2
Medical Long Context Reasoning (MLCR-AA, October 2026 methodology)Health & Medicine13.939.4+25.6
AA-PersonalityBench Expressiveness indexChat & Writing84.860.0−25.0
FrontierMath Tier 4 (v2) - PrivateMath78.0100.0+22.0
FrontierCode 2.0 (Artificial Analysis)Code25.747.1+21.4
Tax Agent BenchFinance & Accounting79.057.9−21.2
Chess Puzzles (task version 2.0)Games59.078.0+19.0
AA-PersonalityBench Conviction indexChat & Writing63.880.9+17.3
Furniture Assembly BenchmarkMultimodal72.589.2+16.7
AA-PersonalityBench Transparency indexChat & Writing21.936.1+14.3
AA-PersonalityBench Imagination indexChat & Writing46.760.4+13.9
Vals Science IndexIndexes55.468.4+13.1
AA-AnalystAgent (pass^5)Professional Work48.861.3+12.5
Legal Research BenchLegal54.842.8−12.0
Artificial Analysis Intelligence Index v5.0.1Indexes50.462.1+11.7
AA-PersonalityBench Warmth indexChat & Writing79.167.8−11.4
CritPt v2 (agentic, avg@5)Science72.983.7+10.9
HLAB-Vals-1.1Legal27.116.3−10.8
Agents' Last Exam v1.1 ScoreAgents69.078.3+9.3
Code Migration v1.1 (Vals Index v2.2 subset)Code74.483.6+9.2
MirrorCode (ML, +Private, 2L)Code58.967.8+8.9
FrontierMath ErdősMath1.510.3+8.8
Vals Finance IndexIndexes66.057.3−8.7
MysteryMechanismScience45.553.2+7.7
Vals Legal IndexIndexes76.869.7−7.1
Finance Agent v2.1Finance & Accounting67.760.7−6.9
FrontierMath Tiers 1-3 (v2) - PrivateMath86.092.6+6.7
Epoch Capabilities Index (ECI)Indexes167174+6.1
Terminal-Bench 4.0Code57.663.1+5.6
AA-Briefcase v1.1 EloAgents14851570+4.7
ARC-AGI-2General QA91.395.8+4.6
BioMysteryBench (Vals Terminus 2)Science76.380.7+4.4
GDP.pdf (Artificial Analysis, All-pass)Professional Work25.629.8+4.2
Terminal-Bench 4.0 (Artificial Analysis)Code55.158.6+3.5
SEC-Bench Pro - release-reported cohortCybersecurity83.486.8+3.3
SimpleQA Verified (Epoch anti-abstention)General QA76.078.5+2.5
ARC-AGI-3 (Semi-Private)General QA97.399.8+2.5
AutomationBench 1.0.6 (Zapier)Agents50.152.4+2.3
AA-Omniscience v1.1 IndexIndexes46.150.0+2.0
Excel Modeling Benchmark v1.1Finance & Accounting74.876.0+1.2
MMMU ProMultimodal88.389.5+1.2
GPQA DiamondGeneral QA96.496.8+0.4
AutomationBench-AA 1.0.6Agents77.577.2−0.3
ARC-AGI-1 Semi-PrivateGeneral QA98.598.50.0
OTIS Mock AIME 2024-2025Math100.0100.00.0

Gemini 4 ArgonvsGPT-6.2 Astranormalised score marginsData Nov 17, 2026Data: Artificial Analysis, Vals AI, Epoch AI and 5 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…