Nanbeige4.1-3B: benchmark results
Nanbeige Lab's 3B reasoning and agentic model, an SFT+RL post-train of its Nanbeige4-3B base aimed at matching much larger Qwen3 models (February 2026). Provider: Nanbeige. Released 2026-02-11. Access: Open.
Aggregate Rating 1464 ± 37, rank #902 of 1635 rated models, from 53 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| AA GPQA Diamond | 84.95 | Accuracy (%) | 78.2 |
| AA-Omniscience Hallucination Rate | 58.1 | Hallucination Rate (%) | 74 |
| OpenEvals - GPQA Diamond | 83.8 | Accuracy (%) | 65.4 |
| AA-Omniscience Index - Law | -34 | Omniscience Index | 53.6 |
| AA Humanity's Last Exam | 10.89 | Accuracy (%) | 52.3 |
| AA-Omniscience Index - Business | -35.5 | Omniscience Index | 47.2 |
| AA-Omniscience Index - Humanities & Social Sciences | -40.4 | Omniscience Index | 43.3 |
| AA Omniscience | -42.5 | Score | 40.8 |
| AA-Omniscience Index - Science, Engineering & Mathematics | -35.5 | Omniscience Index | 38.6 |
| AA-Omniscience Index - Software Engineering (SWE) - Kotlin | -48 | Omniscience Index | 37.7 |
| AA-Omniscience Index - Software Engineering (SWE) - PHP | -46 | Omniscience Index | 35.8 |
| AA-Omniscience Index - Software Engineering (SWE) - R | -66 | Omniscience Index | 35.6 |
Interactive version: theaggregate.ai/model?slug=nanbeige4-1-3b · How It Works · Data refreshed daily, snapshot 2026-10-15.