Atlas

Compare Models

GPT-6.2 Astra vs GPT-6 Astra

GPT-6.2 Astra leads by 1.9 capability points (≈95% interval −1.0 to +4.8): statistically tied · P(higher) ≈ 90%

Wins 48 of 68 shared benchmarks (14 lost, 6 tied)

Capability vs. relative cost
GPT-6.2 AstraGPT-6 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.

GPT-6.2 AstravsGPT-6 Astracapability vs. relative costData Nov 17, 2026Village LLM Atlasatlas.village.dev

Benchmark wins48 – 14 · 68 shared, 6 tied

Comparing matching setups only (2 benchmarks excluded) · suites counted once
← GPT-6.2 AstraGPT-6 Astra →

GPT-6.2 AstravsGPT-6 Astrawins by categoryData Nov 17, 2026Data: Artificial Analysis, Vals AI, Epoch AI and 3 othersVillage LLM Atlasatlas.village.dev

Margins

Each benchmark scaled 0–100% (its declared range, else the field’s range); Δ is GPT-6 Astra minus GPT-6.2 Astra.

AA-PersonalityBench Transparency indexChat & Writing36.194.3+58.8
AA-PersonalityBench Organization indexChat & Writing16.669.5+53.4
AA-PersonalityBench Conviction indexChat & Writing80.944.8−36.5
ExploitBench - Internal Port (June-August 2026)Cybersecurity53.529.0−24.5
KernelGen1P - OpenAI release percentageCode76.966.7−18.7
SecureSWE (OpenAI internal)Cybersecurity65.748.7−17.0
AA-PersonalityBench Imagination indexChat & Writing60.477.2+16.9
EBR-bench v4 (card ban, single agent)Games92.476.2−16.2
Terminal-Bench-Science 0.1Science75.259.5−15.7
AA-PersonalityBench Slop indexChat & Writing1.016.2+15.4
AA-AnalystAgent (pass^5)Professional Work62.551.3−11.3
Mystery Game PuzzlesGames95.084.0−11.0
AutomationBench 1.0.6 (Zapier)Agents52.441.4−11.0
DeepsecBench-AA (Artificial Analysis)Cybersecurity46.836.9−9.9
HLAB-Vals-1.1Legal16.37.1−9.2
Furniture Assembly BenchmarkMultimodal89.280.0−9.2
PatchBench (OpenAI internal)Cybersecurity79.770.9−8.8
AutomationBench-AA 1.0.6Agents77.268.5−8.8
Terminal-Bench Science 0.1 (Vals)Science71.462.9−8.6
Humanity's Last Exam DiamondGeneral QA72.263.8−8.4
MirrorCode (ML, +Private, 2L)Code67.860.0−7.8
AA-PersonalityBench Informativeness indexChat & Writing22.029.6+7.7
FrontierCode 2.0 (Artificial Analysis)Code49.041.4−7.5
AA-PersonalityBench Warmth indexChat & Writing67.875.3+7.5
FrontierMath ErdősMath10.32.9−7.4
SpireClimbGames41.334.0−7.3
LifeVal (GPT-6.2 launch chart)Agents86.179.8−6.3
GDPval v2 (OpenAI internal)Professional Work50.444.2−6.2
Code Migration v1.1 (Vals Index v2.2 subset)Code83.678.0−5.6
Tax Agent BenchFinance & Accounting57.963.3+5.5
MentalHealthBenchSafety53.358.7+5.4
Artificial Analysis Intelligence Index v5.0.1Indexes62.156.9−5.3
Terminal-Bench 4.0 (Artificial Analysis)Code64.159.6−4.5
AA-Briefcase v1.1 EloAgents16441569−4.1
Chess Puzzles (task version 2.0)Games78.074.0−4.0
Epoch Capabilities Index (ECI)Indexes174169−4.0
Artificial Analysis Cyber Index v1Indexes37.333.4−3.8
OpenAI research debugging - internal release scoreCode81.878.0−3.8
Terminal-Bench 4.0Code63.159.6−3.5
Legal Research BenchLegal42.839.4−3.4
Vals Science IndexIndexes68.465.1−3.4
Excel Modeling Benchmark v1.1Finance & Accounting76.072.7−3.3
MMMU ProMultimodal89.586.4−3.1
SimpleQA Verified (Epoch anti-abstention)General QA78.575.6−2.9
AA-Omniscience v1.1 IndexIndexes50.044.6−2.7
FrontierMath Tier 4 (v2) - PrivateMath100.097.6−2.4
GDP.pdf (Artificial Analysis, All-pass)Professional Work29.832.2+2.4
Medical Long Context Reasoning (MLCR-AA, October 2026 methodology)Health & Medicine39.441.7+2.2
HealthBench ProfessionalHealth & Medicine66.068.2+2.2
PostTrainBench Lite - release-reported performanceCode74.877.0+2.2
Vals RSI Index v1.1 (four tasks)Agents29.027.1−1.9
AA-PersonalityBench Expressiveness indexChat & Writing60.058.1−1.9
CWE-Bench-AA (Artificial Analysis)Cybersecurity65.063.3−1.7
Vals Finance IndexIndexes57.355.6−1.7
BioMysteryBench (Vals Terminus 2)Science80.779.3−1.5
SEC-Bench Pro - release-reported cohortCybersecurity86.885.4−1.3
Vals Legal IndexIndexes69.768.4−1.3
GPQA DiamondGeneral QA96.895.8−1.1
FrontierMath Tiers 1-3 (v2) - PrivateMath92.693.7+1.1
ARC-AGI-2General QA95.895.0−0.8
Finance Agent v2.1Finance & Accounting60.760.4−0.3
ARC-AGI-3 (Semi-Private)General QA99.899.9+0.1
ARC-AGI-1 Semi-PrivateGeneral QA98.598.50.0
CritPt v2 (agentic, avg@5)Science83.783.70.0
CyberGym-E2E-AA (Artificial Analysis)Cybersecurity0.00.00.0
ExploitBench - release-reported exploit progressCybersecurity100.0100.00.0
MysteryMechanismScience53.253.20.0
OTIS Mock AIME 2024-2025Math100.0100.00.0

GPT-6.2 AstravsGPT-6 Astranormalised score marginsData Nov 17, 2026Data: Artificial Analysis, Vals AI, Epoch AI and 3 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…