Atlas

Compare Models

Gemini 3.8 Flash vs GPT-6.2 Astra

GPT-6.2 Astra leads by 8.3 capability points (≈95% interval +5.6 to +11.0): a clear lead over Gemini 3.8 Flash · P(higher) ≈ 100%

Wins 43 of 54 shared benchmarks (10 lost, 1 tied)

Capability vs. relative cost
Gemini 3.8 FlashGPT-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 3.8 FlashvsGPT-6.2 Astracapability vs. relative costData Nov 17, 2026Village LLM Atlasatlas.village.dev

Benchmark wins10 – 43 · 54 shared, 1 tied

Comparing matching setups only (10 benchmarks excluded) · suites counted once
← Gemini 3.8 FlashGPT-6.2 Astra →

Gemini 3.8 FlashvsGPT-6.2 Astrawins by categoryData Nov 17, 2026Data: Artificial Analysis, Vals AI, Epoch AI and 2 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 3.8 Flash.

FrontierMath Tier 4 (v2) - PrivateMath22.0100.0+78.0
AA-PersonalityBench Slop indexChat & Writing71.41.0−71.2
Terminal-Bench-Science 0.1Science10.075.2+65.2
ARC-AGI-3 (Semi-Private)General QA35.099.8+64.8
Terminal-Bench Science 0.1 (Vals)Science8.671.4+62.9
AA-PersonalityBench Informativeness indexChat & Writing81.022.0−59.6
Furniture Assembly BenchmarkMultimodal31.789.2+57.5
AA-PersonalityBench Organization indexChat & Writing73.316.6−57.3
Mystery Game PuzzlesGames47.095.0+48.0
Code Migration v1.1 (Vals Index v2.2 subset)Code38.283.6+45.4
Terminal-Bench 4.0Code19.763.1+43.4
Humanity's Last Exam DiamondGeneral QA29.172.2+43.1
Terminal-Bench 4.0 (Artificial Analysis)Code19.760.1+40.4
Vals Science IndexIndexes35.868.4+32.7
Artificial Analysis Intelligence Index v5.0.1Indexes30.862.1+31.3
DeepsecBench-AA (Artificial Analysis)Cybersecurity14.544.1+29.5
MirrorCode (ML, +Private, 2L)Code38.967.8+28.9
FrontierCode 2.0 (Artificial Analysis)Code18.747.1+28.4
FrontierMath Tiers 1-3 (v2) - PrivateMath68.492.6+24.2
Medical Long Context Reasoning (MLCR-AA, October 2026 methodology)Health & Medicine15.639.4+23.9
AutomationBench 1.0.6 (Zapier)Agents29.752.4+22.7
AA-Briefcase v1.1 EloAgents12041570+20.1
Chess Puzzles (task version 2.0)Games58.078.0+20.0
AA-PersonalityBench Imagination indexChat & Writing41.060.4+19.6
AA-PersonalityBench Expressiveness indexChat & Writing79.160.0−19.3
CritPt v2 (agentic, avg@5)Science65.183.7+18.6
BioMysteryBench (Vals Terminus 2)Science62.280.7+18.5
MysteryMechanismScience36.553.2+16.7
AutomationBench-AA 1.0.6Agents60.977.2+16.4
AA-AnalystAgent (pass^5)Professional Work46.361.3+15.0
CWE-Bench-AA (Artificial Analysis)Cybersecurity52.565.8+13.3
AA-Omniscience v1.1 IndexIndexes23.750.0+13.1
AA-PersonalityBench Transparency indexChat & Writing23.836.1+12.4
Epoch Capabilities Index (ECI)Indexes159174+12.2
Vals Index v2.2Indexes56.067.7+11.7
Artificial Analysis Cyber Index v1Indexes25.136.6+11.5
Vibe Code Bench v1.1Code78.789.9+11.2
FrontierMath ErdősMath0.010.3+10.3
Tax Agent BenchFinance & Accounting66.857.9−8.9
SimpleQA Verified (Epoch anti-abstention)General QA69.778.5+8.8
CyberGym-E2E-AA (Artificial Analysis)Cybersecurity8.40.0−8.4
AA-PersonalityBench Warmth indexChat & Writing60.067.8+7.9
GDP.pdf (Artificial Analysis, All-pass)Professional Work22.829.8+7.0
ARC-AGI-2General QA89.295.8+6.7
Finance Agent v2.1Finance & Accounting65.860.7−5.0
MMMU ProMultimodal84.589.5+5.0
HLAB-Vals-1.1Legal12.116.3+4.2
AA-PersonalityBench Conviction indexChat & Writing84.880.9−3.9
Legal Research BenchLegal38.942.8+3.8
Excel Modeling Benchmark v1.1Finance & Accounting72.976.0+3.2
Vals Finance IndexIndexes60.557.3−3.2
OTIS Mock AIME 2024-2025Math98.9100.0+1.1
Vals Legal IndexIndexes69.869.7−0.1
ARC-AGI-1 Semi-PrivateGeneral QA98.598.50.0

Gemini 3.8 FlashvsGPT-6.2 Astranormalised score marginsData Nov 17, 2026Data: Artificial Analysis, Vals AI, Epoch AI and 2 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…