Hy4 preview: benchmark results
Tencent's open-weight Hy4 preview (August 2026), a 770-billion-parameter Mixture-of-Experts model from the Tencent Hy team, activating 49B parameters per token with a built-in speculative-decoding module. Provider: Tencent. Released 2026-08-28. Access: Open.
Aggregate Rating 1837 ± 10, rank #39 of 1640 rated models, from 234 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| LLM Stats (CritPT) | 16.9 | Score (%) | 100 |
| LLM Stats (MathArena Apex) | 74.2 | Score (%) | 100 |
| LLM Stats (SuperChem) | 66.4 | Score (%) | 100 |
| LLM Stats (WideSearch) | 83.9 | Score (%) | 100 |
| 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 |
| ReLE - Language and Instruction Following | 78.8 | Accuracy (%) | 100 |
| STEMBench120 v1.0.0 (Closed Book) - Biology | 2.59 | Mean fractional scientific progress (0–4; 20 tasks/domain; z | 100 |
| STEMBench120 v1.0.0 (Closed Book) - Chemistry | 2.61 | Mean fractional scientific progress (0–4; 20 tasks/domain; z | 100 |
| STEMBench120 v1.0.0 (Closed Book) - Engineering | 2.62 | Mean fractional scientific progress (0–4; 20 tasks/domain; z | 100 |
| STEMBench120 v1.0.0 (Closed Book) - Overall | 2.59 | Mean fractional scientific progress (0–4; 20 tasks/domain; z | 100 |
Interactive version: theaggregate.ai/model?slug=hy4-preview · How It Works · Data refreshed daily, snapshot 2026-10-13.