DeepSeek V4 Pro Preview: benchmark results

Provider: DeepSeek. Released 2026-04-23. Access: Open.

Unified ELO 1683 ± 1, rank #81 of 1392 rated models, from 10 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
BenchmarkList ECI142.96Capability Index (ECI)91.3
MineBench1551Elo Rating51.6
NYT Connections Extended59.9Score (%)49
Creative Writing (Lechmazur)-0.2Mean Score46.8
Multi-turn Debate (Lechmazur)1471.8Bradley-Terry Rating40
Toolathlon55.9Score (self-reported)37
NL2Repo38.5Score (self-reported)30.8
ComplexConstraints28Score (%)28.2
HANDBOOK.md Agents6.9Score (%)20.2
GDPevo43.58base accuracy (%, no-evolution mode)0

Interactive version: theaggregate.ai/model?slug=deepseek-v4-pro-preview · How It Works · Data refreshed daily, snapshot 2026-09-05.