hy4 (High): benchmark results
Provider: Tencent. Released 2026-11-04. Access: API.
Aggregate Rating 1836 ± 22, rank #114 of 2286 rated models, from 24 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| ParseBench - Charts | 70.96 | ChartDataPointMatch Score | 99.2 |
| Mercor APEX | 77.6 | Best Pass@1 (%) | 91.8 |
| ParseBench - Semantic Formatting | 75.41 | Semantic Formatting Score | 91.7 |
| AI Coding Daily (OpenCode) - Laravel Code Quality | 18.05 | Laravel Code Quality (max 20) points, LLM-judged rubric scor | 90.9 |
| ParseBench | 73.72 | Overall Score | 88.4 |
| Multi-turn Debate (Lechmazur) | 1646.5 | Bradley-Terry Rating | 85.3 |
| EnterpriseBench (CoreCraft Agents) | 68.2 | Score (%) | 83.9 |
| Creative Writing (Lechmazur) | 2.05 | Mean Score | 78.8 |
| NYT Connections Extended | 83.5 | Score (%) | 62.2 |
| ComplexConstraints | 66.8 | Score (%) | 60 |
| Tuesday Work Index | 51.1 | Published work index (0–100) | 56.1 |
| ParseBench - Visual Grounding | 54.54 | Element Pass Rate | 53.6 |
Interactive version: theaggregate.ai/model?slug=hy4-high · How It Works · Data refreshed daily, snapshot 2026-11-20.