DeepSeek R1 (Jan '25): benchmark results

Provider: DeepSeek. Released 2025-01-20. Access: Open.

Unified ELO 1582 ± 1, rank #316 of 1392 rated models, from 21 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
CritPt60Accuracy (self-reported)91.9
AA Omniscience - Humanities & Social Sciences31.33Accuracy (%)76.6
AA Omniscience - Health30Accuracy (%)74.7
AA-Omniscience Accuracy30.52Accuracy (%)74.5
AA Omniscience - Business24.7Accuracy (%)73.7
AA Omniscience - Software Engineering (SWE)41.8Accuracy (%)73.7
AA Omniscience - Science, Engineering & Mathematics36.1Accuracy (%)72.9
AA Omniscience - Law19.2Accuracy (%)69.8
AA CritPt0.57Accuracy (%)59
AA Omniscience-32.22Score54.8
AA Long Context Reasoning57.67Accuracy (%)53.5
Artificial Analysis Intelligence Index12.01Intelligence Index52.1

Interactive version: theaggregate.ai/model?slug=deepseek-r1-jan-25 · How It Works · Data refreshed daily, snapshot 2026-09-05.