DeepSeek V4 Pro (High) — benchmark results

DeepSeek V4 Pro evaluated at the high reasoning-effort setting. Provider: DeepSeek. Released 2026-04-23. Access: Open.

Unified ELO 1799 ± 45, rank #142 of 1776 rated models, from 9 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
LisanBench0.23Mean Path Length / Current Maximum86.6
OckBench84Accuracy (%)82.3
ALE-Bench1006.08Performance (Self-Refine x1) (self-reported)76.9
Epoch AI - ECI148.87ECI Score69.2
BenchLM55.5Overall Score60.4
WeirdML46.53Average Score59.1
FutureEval8.66Unified Forecasting Score47.3
Surface Evolver Bench Pass Rate25Pass Rate (%)43.2
Surface Evolver Bench40Mean Score (%)22.7

Interactive version: theaggregate.ai/model?slug=deepseek-v4-pro-high · How the rankings work · Data refreshed daily, snapshot 2026-07-22.