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

BenchmarkScoreMetricPercentile
Nejumi 4 - ARC-AGI-1100Accuracy (%)100
Nejumi 4 - jaster (0-shot) - JSQuAD83.3Character F1 (x100)100
Nejumi 4 - jaster (2-shot) - MMLU-ProX (Japanese)92Exact match (%)100
Nejumi 4 - jaster (2-shot) - ALT English-to-Japanese92.72COMET wmt22 (x100)99.3
Nejumi 4 - jaster (0-shot) - JCommonsenseQA100Exact match (%)96.1
Nejumi 4 - jaster (2-shot) - JCommonsenseQA100Exact match (%)94.7
Nejumi 4 - HLE (Japanese) - Accuracy44.33Accuracy (%)94.3
Nejumi 4 - jaster (0-shot) - MMLU-ProX (Japanese)90Exact match (%)91.8
Nejumi 4 - ARC-AGI-266Accuracy (%)91.5
Nejumi 4 - HLE (Japanese) - Calibration Error49Calibration error (%)90.8
Nejumi 4 - jaster (2-shot) - JSQuAD81.3Character F1 (x100)90.1
Nejumi 4 - jaster (0-shot) - JMMLU94Exact match (%)89.7

Interactive version: theaggregate.ai/model?slug=hy4-preview-high · How It Works · Data refreshed daily, snapshot 2026-10-13.