Tencent
Built for real-world business scenarios, Hy3 features a 295B/21B active MoE architecture, native 256K context support, and three reasoning modes. It enhances coding, long-form comprehension, multi-turn dialogue, and agentic task execution, balancing reliability, efficiency, and cost across both high-frequency interactions and complex workflows.
Artificial Analysis model profileRank #63 across 600
Rank #67 across 266
Rank #64 across 379
Percentile score by analysis domain.
* Cost is inverted: lower input, output, and blended prices rank higher.
Higher bars mean stronger relative placement.
| Metric | Domain | Value | Rank |
|---|---|---|---|
| Artificial Analysis Intelligence Index | overall | 41.2 | #63 |
| Artificial Analysis Coding Index | coding | 58.8 | #67 |
| GPQA | reasoning | 89.7% | #42 |
| Humanity's Last Exam | reasoning | 31.6% | #59 |
| SciCode |
| coding, reasoning |
| 47.6% |
| #62 |
| Output Speed | speed | 72.2 tok/s | #106 |
| Time to First Token | speed | 1.65s | #106 |
| Blended Price | cost | $0.241/M | #64 |
| Input Price | cost | $0.136/M | #56 |
| Output Price | cost | $0.557/M | #78 |
| Value Index | cost, overall | 171.0 | #14 |