DeepSeek V4 Pro (0813) (Non-reasoning): benchmark results

Provider: DeepSeek. Released 2026-08-13. Access: Open.

Aggregate Rating 1647 ± 14, rank #554 of 2245 rated models, from 84 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
AA-Omniscience Accuracy41.45Accuracy (%)78.7
German LLM Index - INCLUDE v1.287.85Exact match, strict final-answer parsing (%)75
German LLM Index - ScaLA Acceptability75.61Matthews correlation coefficient (x100)67.4
German LLM Index - MMLU-ProX v1.180.73Accuracy (%)66.7
AA Omniscience-12.38Score61.3
AA-Briefcase - PDF - Rubric Pass Rate35.34Rubric Pass Rate (%)53.2
AutomationBench-AA - Strict14.61Strict Task Success (%)52.9
AA-Briefcase - DOCX - Rubric Pass Rate34.32Rubric Pass Rate (%)51.9
AA Humanity's Last Exam10.57Accuracy (%)49.3
AA-Omniscience v1.1 Accuracy38.6Accuracy (%)46.5
AA GDPval1065.08ELO45.9
AutomationBench-AA34.91Guardrail-Adjusted Objective Completion (%)45.4

Interactive version: theaggregate.ai/model?slug=deepseek-v4-pro-0813-non-reasoning · How It Works · Data refreshed daily, snapshot 2026-11-16.