Hy4 Preview (High): benchmark results
Provider: Tencent. Released 2026-08-28. Access: Open.
Aggregate Rating 1796 ± 23, rank #149 of 2133 rated models, from 72 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Nejumi 4 - ARC-AGI-1 | 100 | Accuracy (%) | 100 |
| Nejumi 4 - jaster (0-shot) - JSQuAD | 83.3 | Character F1 (x100) | 100 |
| Nejumi 4 - jaster (2-shot) - MMLU-ProX (Japanese) | 92 | Exact match (%) | 100 |
| Nejumi 4 - jaster (2-shot) - ALT English-to-Japanese | 92.72 | COMET wmt22 (x100) | 99.3 |
| Nejumi 4 - jaster (0-shot) - JCommonsenseQA | 100 | Exact match (%) | 96.1 |
| Nejumi 4 - jaster (2-shot) - JCommonsenseQA | 100 | Exact match (%) | 94.7 |
| Nejumi 4 - HLE (Japanese) - Accuracy | 44.33 | Accuracy (%) | 94.3 |
| Nejumi 4 - jaster (0-shot) - MMLU-ProX (Japanese) | 90 | Exact match (%) | 91.8 |
| Nejumi 4 - ARC-AGI-2 | 66 | Accuracy (%) | 91.5 |
| Nejumi 4 - HLE (Japanese) - Calibration Error | 49 | Calibration error (%) | 90.8 |
| Nejumi 4 - jaster (2-shot) - JSQuAD | 81.3 | Character F1 (x100) | 90.1 |
| Nejumi 4 - jaster (0-shot) - JMMLU | 94 | Exact match (%) | 89.7 |
Interactive version: theaggregate.ai/model?slug=hy4-preview-high · How It Works · Data refreshed daily, snapshot 2026-10-13.