Hy3-preview (Reasoning) — benchmark results
Tencent Hunyuan Hy3 preview (295B A21B open MoE) evaluated with reasoning enabled. Provider: Tencent. Released 2026-04-24. Access: Open.
Unified ELO 1769 ± 23, rank #168 of 1776 rated models, from 42 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| LLMEval-Logic Base | 75.3 | Accuracy (%) | 92.3 |
| UGI - Natural Intelligence | 50.11 | NatInt Score | 91 |
| UGI - Writing | 51.12 | Writing Score | 89.4 |
| AA TAU-2 Bench | 92.69 | Accuracy (%) | 88.2 |
| AA GPQA Diamond | 86.67 | Accuracy (%) | 87.3 |
| AA Humanity's Last Exam | 25.53 | Accuracy (%) | 85.2 |
| Wolfram LLM Benchmarking Project | 56.7 | Correct Functionality (%) | 85.2 |
| AA Omniscience - Software Engineering (SWE) - Swift | 64 | Accuracy (%) | 84.9 |
| AA CritPt | 4.57 | Accuracy (%) | 84.5 |
| AA Omniscience - Science, Engineering & Mathematics | 37.3 | Accuracy (%) | 81.2 |
| Artificial Analysis Intelligence Index | 33.58 | Intelligence Index | 80.6 |
| AA Terminal-Bench Hard | 34.09 | Accuracy (%) | 78.8 |
Interactive version: theaggregate.ai/model?slug=hy3-preview-reasoning · How the rankings work · Data refreshed daily, snapshot 2026-07-22.