Atlas

Compare Models

GPT-6.2 Astra vs Muse Ember

GPT-6.2 Astra leads by 6.4 capability points (≈95% interval +3.7 to +9.0): a clear lead over Muse Ember · P(higher) ≈ 100%

Wins 40 of 49 shared benchmarks (8 lost, 1 tied)

Capability vs. relative cost
GPT-6.2 AstraMuse EmberClosed weightsOpen weights

Benchmark wins40 – 8 · 49 shared, 1 tied

Comparing matching setups only (8 benchmarks excluded) · suites counted once
← GPT-6.2 AstraMuse Ember →
Data: Artificial Analysis, Vals AI, Surge AI and 5 othersatlas.village.dev

Margins

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

Terminal-Bench-Science 0.1Science › Multidisciplinary78.120.0−58.1
Humanity's Last Exam DiamondKnowledge & reasoning › General knowledge72.225.1−47.1
TaxCalcBench TY25Professional › Finance, accounting & tax44.02.0−42.0
Code Migration v1.1 (Vals Index v2.2 subset)Software › Software engineering83.647.1−36.5
LMArena Agent Arena Net Improvement (Steering Burden)Tool & computer use › General agents15.21.0−27.8
Artificial Analysis Intelligence Index v5.0.1Composite indexes62.136.0−26.1
Artificial Analysis Engineering IndexScience › Engineering66.641.8−24.9
Terminal-Bench 4.0Software › Software engineering63.139.4−23.7
Terminal-Bench 4.0 (Artificial Analysis)Software › Software engineering64.140.9−23.2
RiemannBench (November 2026 rescoring)Math76.053.6−22.4
DAYJOB: Finance (October 2026 rescoring)Professional › Finance, accounting & tax45.824.2−21.6
CritPt v2 (agentic, avg@5)Science › Physical sciences83.762.3−21.4
AA-Omniscience v1.1 IndexKnowledge & reasoning › General knowledge50.07.7−21.1
Medical Long Context Reasoning (MLCR-AA, October 2026 methodology)Health & medicine39.418.3−21.1
Ramp SWE-Bench v2 (Bash harness)Software › Software engineering75.054.4−20.6
Artificial Analysis Legal IndexProfessional › Legal58.738.5−20.3
Artificial Analysis Economics IndexComposite indexes61.141.0−20.1
LMArena Agent Arena Praise vs ComplaintTool & computer use › General agents37.0-2.7−19.9
EnterpriseBench: CoreCraftTool & computer use › Customer service & operations43.662.1+18.5
FrontierCode 2.0 (Artificial Analysis)Software › Software engineering45.427.2−18.2
DAYJOB: Healthcare (October 2026 rescoring)Health & medicine34.417.2−17.2
Artificial Analysis Finance & Accounting IndexProfessional › Finance, accounting & tax60.944.6−16.3
HANDBOOK.md Agents v1.1Tool & computer use › Customer service & operations45.829.8−16.0
Artificial Analysis Healthcare & Medical IndexHealth & medicine57.041.3−15.7
Artificial Analysis Strategy & Ops IndexProfessional › General knowledge work63.348.7−14.6
GDP.xlsxProfessional › General knowledge work29.715.4−14.3
Tax Agent BenchProfessional › Finance, accounting & tax57.971.4+13.5
DrugDiscoveryBenchScience › Life sciences65.552.4−13.0
Professional Reasoning Benchmark - LegalProfessional › Legal43.155.6+12.5
Chartography (November 2026 rescoring)Perception & media › Documents & charts90.779.3−11.4
ComplexConstraints (November 2026 rescoring)Language & writing › Instruction following69.658.8−10.8
AutomationBench-AA 1.0.6Tool & computer use › Customer service & operations77.268.1−9.2
HLAB-Vals-1.1Professional › Legal16.325.0+8.8
LMArena Agent Arena Confirmed SuccessTool & computer use › General agents12.4-2.0−7.2
Vals Index v2.2Composite indexes67.760.5−7.2
GDP.pdf (Artificial Analysis, All-pass)Professional › General knowledge work29.822.8−7.0
AA-Briefcase v1.1 EloProfessional › General knowledge work16441520−6.8
Legal Research BenchProfessional › Legal42.849.0+6.3
Vibe Code Bench v1.1Software › Frontend & design89.983.7−6.2
Vals Legal IndexProfessional › Legal69.775.7+6.0
SimpleBenchKnowledge & reasoning › General knowledge77.372.7−4.6
LMArena Agent Arena Steering BurdenTool & computer use › General agents14.04.8−4.6
SWE Atlas v1.1 - OverallSoftware › Software engineering60.556.0−4.5
Excel Modeling Benchmark v1.1Professional › Finance, accounting & tax76.072.2−3.9
Vals Finance IndexProfessional › Finance, accounting & tax57.360.9+3.6
Finance Agent v2.1Professional › Finance, accounting & tax60.763.8+3.1
LMArena Agent Arena Bash ReliabilityTool & computer use › General agents7.76.6−0.5
LMArena Agent Arena Tool HallucinationTool & computer use › Tool calling1.20.9−0.1
GDP.pdf (November 2026 rescoring)Professional › General knowledge work49.249.20.0
Data: Artificial Analysis, Vals AI, Surge AI and 5 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…