Hy4 preview
Key facts · verified Oct 27, 2026
Hy4 preview is a Tencent model released on 28 Aug 2026. Its API price is $0.834 per million input tokens and $2.501 per million output tokens ($0.042 cached input (Tencent Cloud, Singapore)). Its context window is 1M tokens. It is not yet ranked on LMArena's text leaderboard. Source: tencentcloud.com.
Tencent’s Hy4 preview, released and open-sourced on 28 August 2026: a mixture-of-experts model with 770 billion total and 49 billion active parameters and a 1-million-token context window, which Tencent calls its most capable model to date. Tencent Cloud serves it with up to 960K input and 64K output tokens, reasoning, structured output, function calling and caching, at $0.834 per million input tokens and $2.501 per million output tokens. The weights are under Apache 2.0. On LMArena’s text leaderboard it has no LMArena rating.
$0.834 / $2.501 per million tokens
1M tokens
Not rated
Not ranked
Key Features
Full Specifications
$0.834 / $2.501 per million tokens | |
|---|---|
| Identity | |
| Developer | Tencent |
| Released | Aug 2026 |
| Status | Preview |
| Licence | Apache 2.0 (open weights) |
| Self-hostable | Yes — tencent/Hy4-preview on Hugging Face |
| Cost | |
| Blended $/1M tokensinput × 0.75 + output × 0.25 | $1.251 / 1M tokens |
| Input price | $0.834 / 1M tokens |
| Output price | $2.501 / 1M tokens |
| Cached input | $0.042 / 1M tokens |
| Capacity | |
| Context window | 1M tokens |
| Max output | 64K tokens |
| Capability | |
| Function calling | Yes |
| Structured output | Yes |
| Extended reasoning | Yes — deep reasoning |
| Access | |
| API | Yes — Tencent Cloud TokenHub (hy4-preview) |