DeepSeek V4.1 Flash (Non-reasoning): benchmark results
Provider: DeepSeek. Released 2026-09-10. Access: Open.
Aggregate Rating 1566 ± 9, rank #845 of 2201 rated models, from 146 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| CrossProfile Paper Natural HiddenBench - All facts | 89.7 | Accuracy (%) | 100 |
| AA-Omniscience Hallucination Rate | 53.99 | Hallucination Rate (%) | 74.6 |
| NTBench - AgentDyn Near Target Harmful ASR@1 | 3.8 | Attack success rate at 1 (%) | 66.7 |
| NTBench - AgentDyn Near Target Harmful ASR@3 | 10.4 | Attack success rate at 3 (%) | 66.7 |
| AA GDPval | 1348.71 | ELO | 66.6 |
| ParseBench - Semantic Formatting | 66.05 | Semantic Formatting Score | 66.2 |
| AA Omniscience | -10.18 | Score | 64.7 |
| AA-Omniscience Accuracy | 28.45 | Accuracy (%) | 64.3 |
| AutomationBench-AA - Finance - Raw Objective Completion | 79.68 | Raw Objective Completion (%) | 63.9 |
| AA-Briefcase - Presentation Elo | 1234.28 | Elo | 58.5 |
| AutomationBench-AA | 47.81 | Guardrail-Adjusted Objective Completion (%) | 56.9 |
| DecisionPivot - Core Accuracy | 53.1 | Score (%) | 56.6 |
Interactive version: theaggregate.ai/model?slug=deepseek-v4-1-flash-non-reasoning · How It Works · Data refreshed daily, snapshot 2026-11-12.