Claude Sonnet 4.6 Expert

Aletheion methodology evaluation · system_id: claude-sonnet-4-6 · audit date: 2026-03-23

78.4
Vendor: Anthropic
Tier: Expert · rank 9
Tasks: 24 total · 18 passed · 6 failed
Methodology version: 1.0.0
Composite score: 0.784 PASS

Cognitive-dimension radar

PerceptionGenerationAttentionLearningMemoryReasoningMetacognitionExecutive functionsProblem solvingSocial cognitionNoveltyOrchestration

Dimension scores

Perception100.0Generation74.7Attention99.2Learning57.0Memory99.0Reasoning89.1Metacognition88.3Executive functions90.0Problem solving63.1Social cognition89.0Novelty66.6Orchestration46.7

Per-dimension breakdown (task-level)

DimensionScoreTasksPer-task scoresResult
perception1.000 2 (2p / 0f)1.00, 1.00PASS
generation0.747 2 (1p / 1f)0.50, 0.87CONDITIONAL
attention0.992 2 (2p / 0f)0.98, 1.00PASS
learning0.570 2 (1p / 1f)0.00, 0.95CONDITIONAL
memory0.990 1 (1p / 0f)0.99PASS
reasoning0.891 3 (3p / 0f)1.00, 0.90, 0.85PASS
metacognition0.883 2 (2p / 0f)0.91, 0.87PASS
executive_functions0.900 1 (1p / 0f)0.90PASS
problem_solving0.631 3 (2p / 1f)0.98, 0.34, 0.66CONDITIONAL
social_cognition0.890 1 (1p / 0f)0.89PASS
novelty0.666 3 (2p / 1f)0.70, 0.68, 0.64CONDITIONAL
orchestration0.467 2 (0p / 2f)0.60, 0.40FAIL

Strongest dimensions

Weakest dimensions

Evidence provenance
Profile: corpus_taas_profiles[system_id=claude-sonnet-4-6] · Audit definition: corpus_agi_audits[system_id=claude-sonnet-4-6] · System profile: agi_evaluation_profiles[system=claude-sonnet-4-6] · Persona runs: agi_complex_evaluations[system.system_id=claude-sonnet-4-6] · Methodology config: agi_audits.config
Per-dimension evidence: /evidence/alitheion-eval-claude-sonnet-4-6

Live API endpoint: /api/v4/site-data/evaluation/claude-sonnet-4-6
Methodology document: /api/v4/site-data/methodology