DeepSeek V4.1 Flash (Max): benchmark results

Provider: DeepSeek. Released 2026-09-10. Access: Open.

Aggregate Rating 1880 ± 6, rank #58 of 2170 rated models, from 545 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
AgenticBBO Frontier - Bigblue1 Chip Placement0.67Normalized anytime/final score (0-1)100
Codeforces Rating (DeepSeek V4.1 report)3471Codeforces rating, DeepSeek V4.1 report, Max effort (self-re100
EnvironmentDuels - Difficulty29.7Score (%)100
Fangcun DeepSeek 1017 Report - Alignment Faking draft training excluded2Native rate (%)100
Fangcun DeepSeek 1017 Report - DeceptionBench L2 single incentive0.33Rate (%)100
Fangcun DeepSeek 1017 Report - DeceptionBench L3 adaptive follow-up4.83Rate (%)100
Fangcun DeepSeek 1017 Report - GDM Oversight Pattern domain review20Native rate (%)100
Fangcun DeepSeek 1017 Report - GDM Oversight Pattern subject tag2Native rate (%)100
Fangcun DeepSeek 1017 Report - GDM Self-Reasoning Message Limit8Successful trials100
Fangcun DeepSeek 1017 Report - GDM Self-Reasoning configuration tasks53.3Accuracy (%)100
Fangcun DeepSeek 1017 Report - INTIMA relationship boundary64.1Rate (%)100
Fangcun DeepSeek 1017 Report - ImpossibleBench unmodified LiveCodeBench96Native rate (%)100

Interactive version: theaggregate.ai/model?slug=deepseek-v4-1-flash-max · How It Works · Data refreshed daily, snapshot 2026-11-05.