Nanbeige3.5-Pro (Thinking): benchmark results
Provider: Nanbeige. Access: API.
Unified ELO 1579 ± 1, rank #757 of 3078 rated models, from 58 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| BFCL V4 - Relevance Detection | 100 | Accuracy (%) | 96.3 |
| OpenCompass Agent - Tool Use (CompassBench 2510) | 72.1 | Score (%) | 88.5 |
| BFCL V4 - Memory | 45.16 | Accuracy (%) | 87.8 |
| OpenCompass Math - College - English (CompassBench 2510) | 56 | Score (%) | 75 |
| BFCL V4 - Web Search | 42 | Accuracy (%) | 73.2 |
| OpenCompass Language - NLP - English (CompassBench 2510) | 94.1 | Score (%) | 71.2 |
| BFCL V4 - Multi Turn | 40 | Accuracy (%) | 70.7 |
| Gorilla API Bench (BFCL) | 47.68 | Overall Accuracy (%) | 69.5 |
| OpenCompass Language - NLP - Chinese (CompassBench 2510) | 81.2 | Score (%) | 67.3 |
| OpenCompass LLM - Language - Chinese (CompassBench 2510) | 84.7 | Score (%) | 65.4 |
| OpenCompass Language - Instruction Following - Chinese (CompassBench 2510) | 76.8 | Score (%) | 65.4 |
| OpenCompass Math - Reasoning Computation - English (CompassBench 2510) | 78.8 | Score (%) | 65.4 |
Interactive version: theaggregate.ai/model?slug=nanbeige3-5-pro-thinking · How It Works · Data refreshed daily, snapshot 2026-09-19.