Atlas

Compare Models

GPT-6.2 Astra vs Muse Spark 1.3

GPT-6.2 Astra leads by 7.5 capability points (≈95% interval +5.0 to +10.0): a clear lead over Muse Spark 1.3 · P(higher) ≈ 100%

Wins 51 of 67 shared benchmarks (16 lost)

Capability vs. relative cost
GPT-6.2 AstraMuse Spark 1.3Closed weightsOpen weights

Benchmark wins51 – 16 · 67 shared

Comparing matching setups only (16 benchmarks excluded) · suites counted once
← GPT-6.2 AstraMuse Spark 1.3 →
Data: Artificial Analysis, Surge AI, Epoch AI and 9 othersatlas.village.dev

Margins

Each benchmark scaled 0–100% (its declared range, else the field’s range); Δ is Muse Spark 1.3 minus GPT-6.2 Astra.

Mystery Game PuzzlesGames95.025.0−70.0
AA-PersonalityBench Informativeness indexLanguage & writing › Style & personality31.3100+69.4
AA-PersonalityBench Slop indexLanguage & writing › Style & personality7.475.3+68.5
Terminal-Bench-Science 0.1Science › Multidisciplinary78.111.0−67.1
Terminal-Bench Science 0.1 (Vals)Science › Multidisciplinary71.410.0−61.4
CyberGym-E2E-AA (Artificial Analysis)Software › Cybersecurity0.059.5+59.5
FrontierMath Tier 4 (v2) - PrivateMath100.046.3−53.7
AA-PersonalityBench Organization indexLanguage & writing › Style & personality21.863.8+42.5
Chess Puzzles (task version 2.0)Games78.037.0−41.0
DeepsecBench-AA (Artificial Analysis)Software › Cybersecurity52.712.5−40.2
AA-3D v1.0 EloSoftware › Frontend & design13181032−38.9
Humanity's Last Exam DiamondKnowledge & reasoning › General knowledge72.233.7−38.5
Terminal-Bench 4.0Software › Software engineering63.124.7−38.4
TaxCalcBench TY25Professional › Finance, accounting & tax44.06.0−38.0
AA-PersonalityBench Imagination indexLanguage & writing › Style & personality63.398.1+35.2
AutomationBench 1.0.6 (Zapier)Tool & computer use › Customer service & operations52.420.7−31.7
Terminal-Bench 4.0 (Artificial Analysis)Software › Software engineering64.133.3−30.8
DAYJOB: Law (November 2026 release)Professional › Legal41.411.8−29.6
AA-PersonalityBench Warmth indexLanguage & writing › Style & personality68.196.2+28.4
LMArena Agent Arena Net Improvement (Steering Burden)Tool & computer use › General agents15.21.1−27.5
Humanity's Sixth SensePerception & media › Spatial & visual reasoning63.237.4−25.8
Artificial Analysis Intelligence Index v5.0.1Composite indexes62.136.6−25.5
sudo L7 (October 2026 rescoring)Software › Software engineering38.312.8−25.5
AA-AnalystAgent (pass^5)Professional › General knowledge work62.537.5−25.0
DAYJOB: AccountingProfessional › Finance, accounting & tax42.018.0−24.0
Medical Long Context Reasoning (MLCR-AA, October 2026 methodology)Health & medicine39.415.6−23.9
RiemannBench (November 2026 rescoring)Math76.053.6−22.4
FrontierCode 2.0 (Artificial Analysis)Software › Software engineering45.423.1−22.3
AutomationBench-AA 1.0.6Tool & computer use › Customer service & operations77.257.9−19.4
DAYJOB: Finance (October 2026 rescoring)Professional › Finance, accounting & tax45.827.0−18.8
LMArena Agent Arena Praise vs ComplaintTool & computer use › General agents37.00.1−18.5
Professional Reasoning Benchmark - LegalProfessional › Legal43.161.6+18.5
EnterpriseBench: CoreCraftTool & computer use › Customer service & operations43.661.9+18.3
FrontierMath Tiers 1-3 (v2) - PrivateMath92.674.4−18.2
MysteryMechanismMath53.236.0−17.1
LiveBench 2.0 (2026-11-06 release)Composite indexes84.968.6−16.3
CritPt v2 (agentic, avg@5)Science › Physical sciences83.768.0−15.7
Tax Agent BenchProfessional › Finance, accounting & tax57.972.4+14.6
LLM Debate BenchmarkLanguage & writing › Language & wordplay16731578−13.7
Gray Swan IPI Attack Success (k=15)Safety & alignment2.415.9−13.5
MirrorCode (ML, +Private, 2L)Software › Software engineering67.854.4−13.3
AA-Omniscience v1.1 IndexKnowledge & reasoning › General knowledge50.023.8−13.1
Legal Research BenchProfessional › Legal42.855.3+12.5
DAYJOB: Healthcare (October 2026 rescoring)Health & medicine34.422.0−12.4
HANDBOOK.md Agents v1.1Tool & computer use › Customer service & operations45.833.9−11.9
Epoch Capabilities Index (ECI)Composite indexes173158−11.7
CWE-Bench-AA (Artificial Analysis)Software › Cybersecurity69.260.8−8.3
GDP.xlsxProfessional › General knowledge work29.722.6−7.1
LMArena Agent Arena Steering BurdenTool & computer use › General agents14.00.8−6.6
Harvey LAB-AA v1.1 (hallucination-gated)Professional › Legal15.08.9−6.1
Chartography (November 2026 rescoring)Perception & media › Documents & charts90.784.6−6.1
AA-PersonalityBench Conviction indexLanguage & writing › Style & personality71.377.2+6.0
DAYJOB: TaxProfessional › Finance, accounting & tax9.23.6−5.6
AA-PersonalityBench Expressiveness indexLanguage & writing › Style & personality66.571.4+5.0
LMArena Agent Arena Confirmed SuccessTool & computer use › General agents12.42.5−5.0
ComplexConstraints (November 2026 rescoring)Language & writing › Instruction following69.664.8−4.8
SimpleBenchKnowledge & reasoning › General knowledge77.381.8+4.5
Vibe Code Bench v1.1Software › Frontend & design89.985.9−4.0
GDP.pdf (November 2026 rescoring)Professional › General knowledge work49.253.0+3.8
Artificial Analysis Cyber Index v1Software › Cybersecurity40.644.3+3.7
AA-Briefcase v1.1 EloProfessional › General knowledge work16441581−3.4
LMArena Agent Arena Bash ReliabilityTool & computer use › General agents7.71.2−3.3
DrugDiscoveryBenchScience › Life sciences65.562.2−3.2
GDP.pdf (Artificial Analysis, All-pass)Professional › General knowledge work29.826.6−3.2
OTIS Mock AIME 2024-2025Math100.099.2−0.8
LMArena Agent Arena Tool HallucinationTool & computer use › Tool calling1.21.0−0.1
AA-PersonalityBench Transparency indexLanguage & writing › Style & personality40.941.0+0.1
Data: Artificial Analysis, Surge AI, Epoch AI and 9 othersatlas.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…