Atlas

Compare Models

Claude Fable 5.1 vs GPT-6.2 Astra

GPT-6.2 Astra leads by 3.0 capability points (≈95% interval +0.3 to +5.7): a clear lead over Claude Fable 5.1 · P(higher) ≈ 98%

Wins 43 of 68 shared benchmarks (24 lost, 1 tied)

Capability vs. relative cost
Claude Fable 5.1GPT-6.2 AstraClosed weightsOpen weights

Benchmark wins24 – 43 · 68 shared, 1 tied

Comparing matching setups only (35 benchmarks excluded) · suites counted once
← Claude Fable 5.1GPT-6.2 Astra →
Data: Artificial Analysis, Surge AI, Epoch AI and 16 othersatlas.village.dev

Margins

Each benchmark scaled 0–100% (its declared range, else the field’s range); Δ is GPT-6.2 Astra minus Claude Fable 5.1.

AA-PersonalityBench Informativeness indexLanguage & writing › Style & personality96.231.3−65.5
AA-PersonalityBench Transparency indexLanguage & writing › Style & personality98.140.9−57.8
SpireClimbGames2.041.3+39.3
AA-PersonalityBench Organization indexLanguage & writing › Style & personality60.021.8−38.6
Mystery Game PuzzlesGames58.095.0+37.0
EBR-bench v4 (card ban, single agent)Games57.192.4+35.2
EnterpriseBench: CoreCraftTool & computer use › Customer service & operations77.443.6−33.8
AA-PersonalityBench Expressiveness indexLanguage & writing › Style & personality98.166.5−31.9
Chess Puzzles (task version 2.0)Games48.078.0+30.0
AutomationBench 1.0.6 (Zapier)Tool & computer use › Customer service & operations22.452.4+30.0
AA-PersonalityBench Conviction indexLanguage & writing › Style & personality10071.3−29.0
AA-PersonalityBench Warmth indexLanguage & writing › Style & personality94.368.1−26.5
MirrorCode (ML, +Private, 2L)Software › Software engineering93.367.8−25.6
Terminal-Bench-Science 0.1Science › Multidisciplinary52.677.1+24.5
AA-3D v1.0 EloSoftware › Frontend & design11391318+24.3
AA-PersonalityBench Imagination indexLanguage & writing › Style & personality86.763.3−23.6
Medical Long Context Reasoning (MLCR-AA, October 2026 methodology)Health & medicine18.339.4+21.1
Tax Agent BenchProfessional › Finance, accounting & tax77.957.9−20.0
Blueprint-Bench 2Perception & media › Spatial & visual reasoning0.40.6+19.5
Furniture Assembly BenchmarkPerception & media › Spatial & visual reasoning70.089.2+19.2
DAYJOB: Law (November 2026 release)Professional › Legal22.541.4+18.9
FrontierCode 2.0 (Artificial Analysis)Software › Software engineering30.347.1+16.8
AA-PersonalityBench Slop indexLanguage & writing › Style & personality23.87.4−16.6
SWE Atlas v1.1 - OverallSoftware › Software engineering76.560.5−16.0
Humanity's Last Exam DiamondKnowledge & reasoning › General knowledge51.366.9+15.6
DAYJOB: AccountingProfessional › Finance, accounting & tax27.242.0+14.8
FrontierMath Tier 4 (v2) - PrivateMath87.8100.0+12.2
Ramp SWE-Bench v2 (Bash harness)Software › Software engineering63.875.0+11.2
Chartography (November 2026 rescoring)Perception & media › Documents & charts79.690.7+11.1
FrontierMath ErdősMath0.010.3+10.3
DAYJOB: Finance (October 2026 rescoring)Professional › Finance, accounting & tax36.545.8+9.3
SimpleBenchKnowledge & reasoning › General knowledge86.677.3−9.3
ComplexConstraints (November 2026 rescoring)Language & writing › Instruction following60.469.6+9.2
Harvey LAB-AA v1.1 (hallucination-gated)Professional › Legal6.415.0+8.6
Professional Reasoning Benchmark - LegalProfessional › Legal51.643.1−8.5
HealthBench Professional (Length-Adjusted)Health & medicine58.166.3+8.2
SimpleQA Verified (Epoch anti-abstention)Knowledge & reasoning › General knowledge70.878.5+7.7
DualEntry Accounting AI BenchmarkProfessional › Finance, accounting & tax85.178.2−6.9
GDP.xlsxProfessional › General knowledge work23.129.7+6.6
sudo L7 (October 2026 rescoring)Software › Software engineering32.238.3+6.1
CritPt v2 (agentic, avg@5)Science › Physical sciences77.783.7+6.0
ARC-AGI-2Knowledge & reasoning › Abstract reasoning90.095.8+5.8
MysteryMechanismMath47.753.2+5.4
AA-AnalystAgent (pass^5)Professional › General knowledge work57.562.5+5.0
RiemannBench (November 2026 rescoring)Math71.276.0+4.8
Epoch Capabilities Index (ECI)Composite indexes167173+4.6
LLM Debate BenchmarkLanguage & writing › Language & wordplay17061673−4.6
LMArena Agent Arena Praise vs ComplaintTool & computer use › General agents28.137.0+4.4
MMMU ProKnowledge & reasoning › General knowledge86.189.5+3.4
GPQA DiamondScience › Multidisciplinary93.797.0+3.3
DAYJOB: Healthcare (October 2026 rescoring)Health & medicine31.234.4+3.2
LMArena Agent Arena Net Improvement (Steering Burden)Tool & computer use › General agents13.715.2+3.0
Dual-Arm Sim (task completion)Tool & computer use › Robotics0.02.5+2.5
FrontierMath Tiers 1-3 (v2) - PrivateMath90.292.6+2.5
TaxCalcBench TY25Professional › Finance, accounting & tax42.044.0+2.0
AA-Briefcase v1.1 EloProfessional › General knowledge work16711644−1.5
Gray Swan IPI Attack Success (k=15)Safety & alignment1.02.4−1.4
HANDBOOK.md Agents v1.1Tool & computer use › Customer service & operations47.145.8−1.3
ARC-AGI-1 Semi-PrivateKnowledge & reasoning › Abstract reasoning97.598.5+1.0
LiveBench 2.0 (2026-11-06 release)Composite indexes84.084.9+0.9
LMArena Agent Arena Confirmed SuccessTool & computer use › General agents14.212.4−0.9
LMArena Agent Arena Bash ReliabilityTool & computer use › General agents9.37.7−0.8
Excel Modeling Benchmark v1.1Professional › Finance, accounting & tax76.776.0−0.7
DAYJOB: TaxProfessional › Finance, accounting & tax9.69.2−0.4
LMArena Agent Arena Steering BurdenTool & computer use › General agents13.514.0+0.2
GDP.pdf (November 2026 rescoring)Professional › General knowledge work49.449.2−0.2
LMArena Agent Arena Tool HallucinationTool & computer use › Tool calling1.01.2+0.1
OTIS Mock AIME 2024-2025Math100.0100.00.0
Data: Artificial Analysis, Surge AI, Epoch AI and 16 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…