Atlas

Compare Models

GPT-6.1 Sol vs GPT-6.2 Astra

GPT-6.2 Astra leads by 2.9 capability points (≈95% interval +0.0 to +5.7): a clear lead over GPT-6.1 Sol · P(higher) ≈ 98%

Wins 46 of 60 shared benchmarks (11 lost, 3 tied)

Capability vs. relative cost
GPT-6.1 SolGPT-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.

GPT-6.1 SolvsGPT-6.2 Astracapability vs. relative costData Nov 17, 2026Village LLM Atlasatlas.village.dev

Benchmark wins11 – 46 · 60 shared, 3 tied

Comparing matching setups only (3 benchmarks excluded) · suites counted once
← GPT-6.1 SolGPT-6.2 Astra →

GPT-6.1 SolvsGPT-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 GPT-6.1 Sol.

AA-PersonalityBench Organization indexChat & Writing81.016.6−65.0
AA-PersonalityBench Transparency indexChat & Writing92.436.1−56.9
EBR-bench v4 (card ban, single agent)Games54.392.4+38.1
ExploitBench - Internal Port (June-August 2026)Cybersecurity18.053.5+35.5
AA-PersonalityBench Conviction indexChat & Writing54.380.9+26.9
Terminal-Bench-Science 0.1Science58.178.1+20.0
Terminal-Bench Science 0.1 (Vals)Science52.971.4+18.6
SpireClimbGames23.841.3+17.5
SecureSWE (OpenAI internal)Cybersecurity49.765.7+16.0
Mystery Game PuzzlesGames80.095.0+15.0
AA-PersonalityBench Slop indexChat & Writing14.31.0−13.5
Humanity's Last Exam DiamondGeneral QA58.872.2+13.4
AA-AnalystAgent (pass^5)Professional Work50.062.5+12.5
AutomationBench-AA 1.0.6Agents66.677.2+10.7
FrontierCode 2.0 (Artificial Analysis)Code38.549.0+10.4
Chess Puzzles (task version 2.0)Games68.078.0+10.0
HLAB-Vals-1.1Legal6.316.3+10.0
AA-PersonalityBench Expressiveness indexChat & Writing50.560.0+9.6
Furniture Assembly BenchmarkMultimodal80.089.2+9.2
GDPval v2 (OpenAI internal)Professional Work41.550.4+8.9
Vals Science IndexIndexes59.668.4+8.8
PatchBench (OpenAI internal)Cybersecurity71.079.7+8.7
Code Migration v1.1 (Vals Index v2.2 subset)Code75.183.6+8.5
Terminal-Bench 4.0 (Artificial Analysis)Code56.164.1+8.1
Terminal-Bench 4.0Code55.163.1+8.1
Artificial Analysis Intelligence Index v5.0.1Indexes54.662.1+7.5
MysteryMechanismScience46.453.2+6.8
MirrorCode (ML, +Private, 2L)Code61.167.8+6.7
LifeVal (GPT-6.2 launch chart)Agents79.586.1+6.6
FrontierMath ErdősMath4.410.3+5.9
Agents' Last Exam v1.1 ScoreAgents72.578.3+5.9
AA-PersonalityBench Informativeness indexChat & Writing27.722.0−5.8
Excel Modeling Benchmark v1.1Finance & Accounting71.176.0+4.9
AA-Briefcase v1.1 EloAgents15551644+4.9
MentalHealthBenchSafety57.953.3−4.6
SimpleQA Verified (Epoch anti-abstention)General QA73.978.5+4.6
Tax Agent BenchFinance & Accounting62.357.9−4.5
Epoch Capabilities Index (ECI)Indexes169174+4.4
MMMU ProMultimodal85.189.5+4.4
Legal Research BenchLegal38.542.8+4.3
AA-Omniscience v1.1 IndexIndexes42.450.0+3.8
Vals Index v2.2Indexes63.967.7+3.7
ARC-AGI-3 (Semi-Private)General QA96.4100.0+3.6
Vals RSI Index v1.1 (four tasks)Agents26.629.0+2.4
GDP.pdf (Artificial Analysis, All-pass)Professional Work32.029.8−2.2
AA-PersonalityBench Warmth indexChat & Writing65.767.8+2.1
ARC-AGI-2General QA94.295.8+1.7
AA-PersonalityBench Imagination indexChat & Writing61.960.4−1.5
Vals Finance IndexIndexes55.857.3+1.5
HealthBench ProfessionalHealth & Medicine67.266.0−1.2
CritPt v2 (agentic, avg@5)Science82.683.7+1.1
Medical Long Context Reasoning (MLCR-AA, October 2026 methodology)Health & Medicine38.339.4+1.1
BioMysteryBench (Vals Terminus 2)Science79.680.7+1.1
Finance Agent v2.1Finance & Accounting61.860.7−1.1
FrontierMath Tiers 1-3 (v2) - PrivateMath93.792.6−1.1
Vibe Code Bench v1.1Code88.989.9+1.0
Vals Legal IndexIndexes68.769.7+1.0
ARC-AGI-1 Semi-PrivateGeneral QA98.598.50.0
FrontierMath Tier 4 (v2) - PrivateMath100.0100.00.0
OTIS Mock AIME 2024-2025Math100.0100.00.0

GPT-6.1 SolvsGPT-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…