Nanbeige4.3-33B: benchmark results
Provider: Nanbeige. Released 2026-10-24. Access: Open.
Aggregate Rating 1737 ± 27, rank #135 of 1635 rated models, from 85 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Nanbeige4.3 Release - AA-LCR | 84 | Accuracy (%) | 100 |
| Nanbeige4.3 Release - AIME 2026 | 95.2 | Accuracy (%) | 100 |
| Nanbeige4.3 Release - Automation Bench v1.0.6 (600 public tasks) | 54.2 | Score (%) | 100 |
| Nanbeige4.3 Release - Claw-Eval pass^3 (General subset) | 78.3 | pass^3 (%) | 100 |
| Nanbeige4.3 Release - DeepPlanning | 63.3 | Score (%) | 100 |
| Nanbeige4.3 Release - DeepSWE (mini-swe-agent) | 65.5 | Score (%) | 100 |
| Nanbeige4.3 Release - EnterpriseOps-Gym | 41 | Score (%) | 100 |
| Nanbeige4.3 Release - JobBench (Claude Code + search) | 44.8 | Score (%) | 100 |
| Nanbeige4.3 Release - NL2Repo (Claude Code) | 53 | Score (%) | 100 |
| Nanbeige4.3 Release - SWE-Atlas-QnA (GPT-6 Sol judge) | 50.8 | Score (%) | 100 |
| Nanbeige4.3 Release - SciCode | 52.6 | Score (%) | 100 |
| Nanbeige4.3 Release - SkillsBench | 61.1 | Score (%) | 100 |
Interactive version: theaggregate.ai/model?slug=nanbeige4-3-33b · How It Works · Data refreshed daily, snapshot 2026-11-05.