DeepSeek V3.1 Terminus (Thinking): benchmark results
Provider: DeepSeek. Released 2025-09-22. Access: Open.
Aggregate Rating 1654 ± 17, rank #511 of 2170 rated models, from 58 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| LiveFMBench - 2025 | 13.64 | Pass@1 (%) after the faithfulness check on LiveFMBench-2025 | 92.9 |
| LiveFMBench - Pre2025 | 30.52 | Pass@1 (%) after the faithfulness check on LiveFMBench-Pre20 | 85.7 |
| Chatbot Arena (Text - Legal & Government) | 1461 | Arena Score | 80.7 |
| Chatbot Arena (Text - Medicine & Healthcare) | 1466 | Arena Score | 78.8 |
| SuperCLUE General (September 2025) - Agent | 75.83 | Score | 77.8 |
| SuperCLUE General (September 2025) - Science Reasoning | 47.41 | Score | 77.8 |
| AGI-Eval Community - Learning (Chinese) | 82.86 | Accuracy (%) | 76.4 |
| AGI-Eval Community - Interaction (English) | 92.49 | Accuracy (%) | 75.4 |
| AGI-Eval Community - General Reasoning | 97.02 | Accuracy (%) | 75 |
| Chatbot Arena (Text - Longer Query) | 1444 | Arena Score | 74.7 |
| Chatbot Arena (Text - Writing, Literature & Language) | 1408 | Arena Score | 73.8 |
| AGI-Eval Community - Learning | 87.98 | Accuracy (%) | 72.9 |
Interactive version: theaggregate.ai/model?slug=deepseek-v3-1-terminus-thinking · How It Works · Data refreshed daily, snapshot 2026-11-05.