Hy3-preview: benchmark results
Tencent Hunyuan Hy3 preview (295B A21B open MoE). Provider: Tencent. Released 2026-04-23. Access: Open.
Aggregate Rating 1723 ± 8, rank #157 of 1656 rated models, from 271 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| CLBench Life - Personal Information Fragments | 20 | Task solving rate (%) on the 45 personal-information-fragmen | 100 |
| VEHBench - Corrupted-State Recovery | 47.4 | P3 success (%; share of corrupted or misleading design traje | 100 |
| AGI-Eval Community - General Reasoning | 97.68 | Accuracy (%) | 98.9 |
| CLBench - Empirical Discovery & Simulation | 19.6 | Solving Rate (%) | 97.1 |
| AA Omniscience - Software Engineering (SWE) - Go | 46 | Accuracy (%) | 95.6 |
| CLBench - Domain Knowledge Reasoning | 25.2 | Solving Rate (%) | 94.3 |
| Trip+ - Intent Preservation | 76.4 | Intent preservation (0-100), a rate scaled by 100: whether e | 94.1 |
| Trip+ - Response Mode Accuracy | 89.3 | Response-mode accuracy (0-100), a rate scaled by 100: whethe | 94.1 |
| AA Omniscience - Software Engineering (SWE) - Julia | 33.33 | Accuracy (%) | 92.8 |
| LLMEval-Logic Base | 75.3 | Accuracy (%) | 92.3 |
| CLBench | 22.8 | Solving Rate (%) | 91.4 |
| OpenCompass Math - College (CompassBench 2604) | 81.3 | Score (%) | 90.9 |
Interactive version: theaggregate.ai/model?slug=hy3-preview · How It Works · Data refreshed daily, snapshot 2026-10-20.