DeepSeek V4 Flash (0731) — benchmark results

Provider: DeepSeek. Released 2026-07-31. Access: Open.

Unified ELO 1861 ± 18, rank #101 of 1839 rated models, from 38 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
AI Chess Leaderboard (Reasoning)1645Elo96.7
LLM Stats Score48.49LLM Stats Score (conservative rating)96.2
PM-LLM-Benchmark37.7Score95.8
AI Chess Leaderboard (Continuation)1409Elo94
Vals AI SWE-bench Verified88.8Resolved (%)93.4
LLM Stats (Toolathlon)70.3Score (%)93.3
Vals AI LiveCodeBench87.26Accuracy (%)92.5
LLM Stats (NL2Repo)54.2Score (%)92.3
Vals AI Vibe Code Bench74.74Accuracy (%)90.9
Vals AI GPQA89.9Accuracy (%)84.4
Vals AI Harvey Legal Agent Bench8.33Accuracy (%)82.8
Vals AI CyberBench77.45Accuracy (%)81

Interactive version: theaggregate.ai/model?slug=deepseek-v4-flash-0731 · How It Works · Data refreshed daily, snapshot 2026-08-05.