DeepSeek V3.1: benchmark results
Switches between direct answers and extended reasoning through its chat template. Additional training targets tool use and agents, including coding and search workflows. Provider: DeepSeek. Released 2025-08-21. Access: Open.
Aggregate Rating 1672 ± 4, rank #264 of 1656 rated models, from 1137 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| CFAgentBench | 67 | pass^1 (%) | 100 |
| ClinTraceBench - Attribution Detection | 93.8 | Accuracy (%; task T3, controlled attribution detection: whet | 100 |
| ClinTraceBench - Cross-Patient Comparison | 54.3 | Accuracy (%; task T6, cross-patient comparison: number of di | 100 |
| CollabBench Cook - Empathy | 2.67 | Empathy score (out of 5): empathy (awareness of and adaptati | 100 |
| CollabBench Cook - Game Score | 136.53 | Final game score of the two-agent kitchen team (completed di | 100 |
| CollabBench Cook - Helpfulness | 1.79 | Helpfulness score (out of 5): helpfulness (task focus, proac | 100 |
| CollabBench Cook - Trustfulness | 2.97 | Trustfulness score (out of 5): trustfulness (reliable commit | 100 |
| EstGraph - Node Count, BA Medium (Simple Walk) | 26.87 | Median relative error of the node-count estimate (%; 100 syn | 100 |
| EstGraph - Node Count, ER Medium (MH Walk) | 4.5 | Median relative error of the node-count estimate (%; 100 syn | 100 |
| EstGraph - Node Count, ER Medium (Simple Walk) | 6 | Median relative error of the node-count estimate (%; 100 syn | 100 |
| EstGraph - Node Count, GRP Medium (Simple Walk) | 6.4 | Median relative error of the node-count estimate (%; 100 syn | 100 |
| Evals for Every Language - Danish (da) | 80.27 | Average Score (%) | 100 |
Interactive version: theaggregate.ai/model?slug=deepseek-v3-1 · How It Works · Data refreshed daily, snapshot 2026-10-20.