Atlas

Compare Models

Claude Sonnet 5.5 vs GPT-6.2 Astra

GPT-6.2 Astra leads by 2.5 capability points (≈95% interval −0.4 to +5.4): statistically tied · P(higher) ≈ 96%

Wins 35 of 47 shared benchmarks (11 lost, 1 tied)

Capability vs. relative cost
Claude Sonnet 5.5GPT-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.

Claude Sonnet 5.5vsGPT-6.2 Astracapability vs. relative costData Nov 17, 2026Village LLM Atlasatlas.village.dev

Benchmark wins11 – 35 · 47 shared, 1 tied

Comparing matching setups only (3 benchmarks excluded) · suites counted once
← Claude Sonnet 5.5GPT-6.2 Astra →

Claude Sonnet 5.5vsGPT-6.2 Astrawins by categoryData Nov 17, 2026Data: Artificial Analysis, Epoch AI, Vals AI and 4 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 Claude Sonnet 5.5.

AA-PersonalityBench Organization indexChat & Writing84.816.6−68.8
AA-PersonalityBench Transparency indexChat & Writing10036.1−64.6
AA-PersonalityBench Informativeness indexChat & Writing84.822.0−63.5
SpireClimbGames1.341.3+40.0
SimpleQA Verified (Epoch anti-abstention)General QA46.578.5+32.0
MirrorCode (ML, +Private, 2L)Code97.867.8−30.0
Mystery Game PuzzlesGames65.095.0+30.0
ARC-AGI-3 (Semi-Private)General QA72.4100.0+27.6
AA-PersonalityBench Slop indexChat & Writing27.71.0−26.9
Terminal-Bench-Science 0.1Science53.378.1+24.8
AA-PersonalityBench Imagination indexChat & Writing84.860.4−24.6
Terminal-Bench Science 0.1 (Vals)Science47.171.4+24.3
AA-PersonalityBench Conviction indexChat & Writing58.180.9+23.1
FrontierMath Tier 4 (v2) - PrivateMath80.5100.0+19.5
Chess Puzzles (task version 2.0)Games59.078.0+19.0
Humanity's Last Exam DiamondGeneral QA53.372.2+18.9
Agents' Last Exam v1.1 ScoreAgents60.478.3+17.9
Furniture Assembly BenchmarkMultimodal75.089.2+14.2
AA-PersonalityBench Expressiveness indexChat & Writing73.360.0−13.5
ARC-AGI-2General QA83.995.8+11.9
AA-Omniscience v1.1 IndexIndexes30.550.0+9.8
AA-Briefcase v1.1 EloAgents18151644−9.4
Artificial Analysis Engineering IndexIndexes58.566.6+8.1
CritPt v2 (agentic, avg@5)Science75.783.7+8.0
Artificial Analysis Intelligence Index v5.0.1Indexes54.562.1+7.6
FrontierMath ErdősMath2.910.3+7.4
Artificial Analysis Economics IndexIndexes54.161.1+7.0
Artificial Analysis Legal IndexIndexes52.258.7+6.6
Epoch Capabilities Index (ECI)Indexes167174+6.0
HLAB-Vals-1.1Legal10.416.3+5.8
AA-PersonalityBench Warmth indexChat & Writing73.367.8−5.6
AutomationBench-AA 1.0.6Agents71.877.2+5.5
Artificial Analysis Strategy & Ops IndexIndexes58.163.3+5.2
Artificial Analysis Finance & Accounting IndexIndexes55.860.9+5.1
AA-AnalystAgent (pass^5)Professional Work57.562.5+5.0
MysteryMechanismScience49.153.2+4.1
GDP.pdf (Artificial Analysis, All-pass)Professional Work25.829.8+4.0
FrontierMath Tiers 1-3 (v2) - PrivateMath88.892.6+3.9
FrontierCode 2.0 (Artificial Analysis)Code41.745.4+3.7
MMMU ProMultimodal85.889.5+3.7
Artificial Analysis Healthcare & Medical IndexIndexes54.257.0+2.8
Vibe Code Bench v1.1Code92.489.9−2.5
Medical Long Context Reasoning (MLCR-AA, October 2026 methodology)Health & Medicine41.739.4−2.2
ARC-AGI-1 Semi-PrivateGeneral QA96.598.5+2.0
Terminal-Bench 4.0 (Artificial Analysis)Code63.664.1+0.5
Excel Modeling Benchmark v1.1Finance & Accounting75.976.0+0.1
OTIS Mock AIME 2024-2025Math100.0100.00.0

Claude Sonnet 5.5vsGPT-6.2 Astranormalised score marginsData Nov 17, 2026Data: Artificial Analysis, Epoch AI, Vals AI and 4 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…