Hy3-preview (Non-reasoning): benchmark results
Tencent Hunyuan Hy3 preview (295B A21B open MoE) evaluated with reasoning disabled. Provider: Tencent. Released 2026-04-24. Access: Open.
Unified ELO 1578 ± 1, rank #524 of 1761 rated models, from 40 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| AA Omniscience - Software Engineering (SWE) - Go | 46 | Accuracy (%) | 95.6 |
| AA Omniscience - Software Engineering (SWE) - Julia | 33.33 | Accuracy (%) | 92.8 |
| AA Omniscience - Software Engineering (SWE) - TypeScript | 42.22 | Accuracy (%) | 86.9 |
| UGI - Natural Intelligence | 37.5 | NatInt Score | 85.4 |
| AA Omniscience - Software Engineering (SWE) - Dart | 30 | Accuracy (%) | 85.2 |
| AA Omniscience - Software Engineering (SWE) - PHP | 40 | Accuracy (%) | 85.2 |
| AA Omniscience - Software Engineering (SWE) - JavaScript | 40 | Accuracy (%) | 81.6 |
| AA Omniscience - Software Engineering (SWE) - R | 22 | Accuracy (%) | 81.4 |
| CritPt | 30 | Accuracy (self-reported) | 80.8 |
| AA Omniscience - Software Engineering (SWE) - Kotlin | 28 | Accuracy (%) | 80.5 |
| UGI Leaderboard | 45.07 | UGI Score | 80.4 |
| AA Omniscience - Software Engineering (SWE) - C | 48 | Accuracy (%) | 78.3 |
Interactive version: theaggregate.ai/model?slug=hy3-preview-non-reasoning · How It Works · Data refreshed daily, snapshot 2026-09-05.