DeepSeek-R1 Competent

Aletheion methodology evaluation · system_id: deepseek-r1 · audit date: 2026-03-23

Vendor: DeepSeek

Tier: Competent · rank 51

Tasks: 24 total · 19 passed · 5 failed

Methodology version: 1.0.0

Composite score: 0.782 PASS

Cognitive-dimension radar

Dimension scores

Per-dimension breakdown (task-level)

Dimension	Score	Tasks	Per-task scores	Result
perception	1.000	2 (2p / 0f)	1.00, 1.00	PASS
generation	0.817	2 (2p / 0f)	0.85, 0.80	PASS
attention	0.992	2 (2p / 0f)	0.98, 1.00	PASS
learning	0.570	2 (1p / 1f)	0.00, 0.95	CONDITIONAL
memory	0.980	1 (1p / 0f)	0.98	PASS
reasoning	0.904	3 (3p / 0f)	1.00, 0.95, 0.85	PASS
metacognition	0.680	2 (1p / 1f)	0.92, 0.56	CONDITIONAL
executive_functions	0.870	1 (1p / 0f)	0.87	PASS
problem_solving	0.604	3 (2p / 1f)	0.98, 0.20, 0.68	CONDITIONAL
social_cognition	0.900	1 (1p / 0f)	0.90	PASS
novelty	0.725	3 (2p / 1f)	0.65, 0.70, 0.78	CONDITIONAL
orchestration	0.607	2 (1p / 1f)	0.70, 0.56	CONDITIONAL

Strongest dimensions

perception — 100.000
attention — 99.200
memory — 98.000

Weakest dimensions

learning — 57.000
problem solving — 60.400
orchestration — 60.700

Evidence provenance

Profile: corpus_taas_profiles[system_id=deepseek-r1] · Audit definition: corpus_agi_audits[system_id=deepseek-r1] · System profile: agi_evaluation_profiles[system=deepseek-r1] · Persona runs: agi_complex_evaluations[system.system_id=deepseek-r1] · Methodology config: agi_audits.config
Per-dimension evidence: /evidence/alitheion-eval-deepseek-r1

Live API endpoint: /api/v4/site-data/evaluation/deepseek-r1
Methodology document: /api/v4/site-data/methodology