DeepSeek V4.1 Flash (1017) (High): benchmark results
Provider: DeepSeek. Released 2026-10-17. Access: Open.
Aggregate Rating 1843 ± 21, rank #83 of 2176 rated models, from 14 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| LisanBench | 0.4 | Mean Path Length / Current Maximum | 94.8 |
| AI Coding Daily (OpenCode) - Laravel Code Quality | 17.98 | Laravel Code Quality (max 20) points, LLM-judged rubric scor | 90.5 |
| ARC-AGI-1 | 94.5 | Accuracy (%) | 85.8 |
| AI Coding Daily (OpenCode) - Total | 55.21 | Total points (max 70) | 81 |
| DuelLab Overall | 66.72 | GameBench2 normalized score (0–100) | 80.8 |
| ARC-AGI-2 | 68.61 | Accuracy (%) | 78.3 |
| Multi-turn Debate (Lechmazur) | 1593 | Bradley-Terry Rating | 71.7 |
| CursorBench 4.0 | 45.6 | Score (%) | 68.2 |
| Argo-Bench | 25.08 | Mean task score (0-100) over 210 enterprise data-science tas | 56.7 |
| Argo-Bench - Solved | 11.48 | Share of the 210 tasks solved (%), a task counting as solved | 51.7 |
| o11y-bench - Pass^3 | 75 | Tasks passed on all three attempts, Pass^3 (%) | 45.4 |
| AI Coding Daily (OpenCode) - React-TS Code Quality | 16.83 | React-TS Code Quality (max 20) points, LLM-judged rubric sco | 42.9 |
Interactive version: theaggregate.ai/model?slug=deepseek-v4-1-flash-1017-high · How It Works · Data refreshed daily, snapshot 2026-10-28.