# Hy3

> Hy3 is a 295B-parameter Mixture-of-Experts model from Tencent (21B active, 192 experts with top-8 routing) built for reasoning, agentic work

Hy3 has a SharkScore of 43.3/100, ranked #79 of 608 in Models (updated 2026-11-19).

- Website: https://openrouter.ai/tencent/hy3 (openrouter.ai)
- Kind: Model · Pricing: Open source
- Maker: Tencent
- Categories: Models, Model APIs & inference
- Verified by the maker: no

## Why it scores 43.3

| Part | Score (0–100) | Weight |
|---|---|---|
| Adoption | 33 | 30% |
| User reviews | no data | 25% |
| Maintenance | 28 | 20% |
| Trust & transparency | 80 | 15% |
| Momentum | no data | 10% |

Evidence covers 87% of the score's weight. Compared with other listings in Models.

## Evidence

| Signal | Value |
|---|---|
| Traffic rank (Tranco) | — |
| GitHub stars | — |
| Stars, last 90 days | — |
| Downloads, 30 days | 981.1K |
| Reviews | none yet |
| Last update | 121 days ago |
| Uptime, 30 days | — |
| Public pricing / Free tier / API / MCP | yes / yes / yes / no |

## RankShark review

Automated evaluation by RankShark · 2026-11-18 · method review-1.0 · https://rankshark.pro/how-we-review
We were not paid for this review.

As of 2026-11-18, Hy3 is Tencent’s text-generation MoE model; the OpenRouter listing shows 262,144 tokens of context and prices of $0.132 per million input tokens and $0.528 per million output tokens. [E1, E2]

Hy3 is a Tencent 295B-parameter mixture-of-experts text-generation model with 21B active parameters. Tencent publishes model weights and instructions for serving it with vLLM or SGLang. [E1]

### Pros

- The OpenRouter listing reports a context window of 262,144 tokens.
- The vendor provides model weights and deployment instructions for vLLM and SGLang.
- The vendor provides a finetuning pipeline and says Hy3 supports GRPO reinforcement-learning training with verl.

### Cons

- Tencent’s self-hosting guidance recommends 8 GPUs and GPUs with larger memory capacity, which may limit deployment options.
- The OpenRouter listing identifies text as the model’s input and output modality.
- No free plan or free trial stated on the pages we checked.
- Whether customer data is used for model training is not stated on the pages we checked.

### Pricing

| Plan | Price | Limits |
|---|---|---|
| OpenRouter inference | $0.132 per million input tokens; $0.528 per million output tokens | OpenRouter lists a 262,144-token context window and a maximum of 128,000 completion tokens. |

Checked 2026-11-18

### Sources

- https://huggingface.co/tencent/Hy3 (fetched 2026-11-18)
- https://openrouter.ai/tencent/hy3 (fetched 2026-11-18)

## Top alternatives

1. [Qwen3.8 27B](https://rankshark.pro/tools/qwen38-27b) — SharkScore 67.1. Qwen3.8 27B is an open-weight dense vision-language model from Qwen.
2. [Qwen3.6 35B A3B](https://rankshark.pro/tools/qwen36-35b-a3b) — SharkScore 65.5. Qwen3.6-35B-A3B is an open-weight multimodal model from Alibaba Cloud with 35 billion total parameters and 3 billion active parameters per t
3. [Qwen3.8 Flash](https://rankshark.pro/tools/qwen38-flash) — SharkScore 64.7. Qwen3.8 Flash is a multimodal reasoning model from Alibaba.
4. [Qwen3.6 27B](https://rankshark.pro/tools/qwen36-27b) — SharkScore 64.3. Qwen3.6 27B is a dense 27-billion-parameter language model from the Qwen Team at Alibaba, released in April 2026.
5. [gpt-oss](https://rankshark.pro/tools/gpt-oss) — SharkScore 63.5. OpenAI's open-weight reasoning models.

---
Source: https://rankshark.pro/tools/hy3 · Method v2026-11-18.2: https://rankshark.pro/how-ranking-works · Payment never changes a rank.
