Skip to content
RankShark

Hy4 preview

Tencent: Hy4 preview is a mixture-of-experts model from Tencent, with 49B active parameters out of 770B total.

ModelOpen sourceby Tencentsince 2026
34.5

SharkScore

#161 in Models

Why it scores 34.5

Hy4 preview has a SharkScore of 34.5/100, ranked #161 of 607 in Models (updated 2026-11-19).

Each part compares Hy4 preview with other listings in Models. Evidence covers 87% of the score's weight.

Adoption
1 · 30%
User reviews
no data · 25%
Maintenance
43 · 20%
Trust & transparency
80 · 15%
Momentum
no data · 10%

Paying never changes this score. How ranking works · method v2026-11-18.2

RankShark review of Hy4 preview

Automated evaluation by RankShark · 18 Nov 2026 · method v1.0 · How we review

We were not paid for this review.

As of 2026-11-18, Hy4 preview is Tencent’s text-only MoE model with a 1,048,576-token context window and usage-based OpenRouter pricing. [E1][E2]

Hy4 preview is a Tencent instruction model with 770B total parameters and 49B activated per token. It supports text input and output and can be deployed with vLLM or SGLang. [E1][E2]

Best for

Not for

  • Teams that need image or audio input; the listed input modality is text. · OpenRouter

Key features

  • Mixture-of-Experts model. The vendor says Hy4 preview has 770B total parameters, with 49B activated per token. · model card
  • Long context. The vendor lists a 1M-token context length; OpenRouter lists 1,048,576 tokens. · model card, OpenRouter
  • Task focus. The vendor says it targets software engineering, office and analysis work, game development, and scientific research. · model card
  • Local serving. The vendor says it can be deployed with vLLM or SGLang and called through an OpenAI-compatible API. · model card
  • Fine-tuning. The vendor says Hy4 preview has a complete model fine-tuning pipeline. · model card

Pros

  • The listed context window is 1,048,576 tokens. · OpenRouter
  • The vendor provides deployment instructions for vLLM and SGLang, including an OpenAI-compatible API example. · model card
  • The vendor says a fine-tuning pipeline is available. · model card

Cons

  • The vendor describes this as an early version and says it can spend longer than necessary reasoning through complex tasks and over-verify its work. · model card
  • OpenRouter lists text as the model’s only input and output modality. · OpenRouter
  • 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

PlanPriceLimits
OpenRouter usage$0.834 per million input tokens; $2.501 per million output tokens; cached input $0.042 per million tokens.OpenRouter lists a 1,048,576-token context window and up to 64,000 completion tokens.

Checked 18 Nov 2026

Free plan
Not stated (checked 18 Nov 2026)
Free trial
Not stated (checked 18 Nov 2026)
Trains on customer data
Not stated (checked 18 Nov 2026)
Data retention
Not stated (checked 18 Nov 2026)
Data regions
Not stated (checked 18 Nov 2026)
Data processing agreement
Not stated (checked 18 Nov 2026)
AI models used
Not stated (checked 18 Nov 2026)
API
OpenRouter lists chat completions, completions, messages, and responses endpoints. The vendor also documents an OpenAI-compatible API for local deployment. · model card, OpenRouter
SDKs
The vendor gives a Python example using the OpenAI client library. · model card
Webhooks
Not stated (checked 18 Nov 2026)
Platforms
The vendor says its SGLang image supports x86 and Arm. [E1] · model card
Security claims
No security certification stated (checked 18 Nov 2026)

What we measured

Context window (OpenRouter listing) · now
1,048,576 tokens
Price per million tokens (OpenRouter listing) · now
$0.834 in, $2.501 out
Package downloads · 30 days
11,299
How we tested · sources
  • Hourly uptime and response-time checks of the product's URL.
  • Repository, release and download data from GitHub, npm, PyPI and Hugging Face where they exist.
  • We did not sign up, pay, or use the product as a customer.
  • 18 Nov 2026: Plans or prices changed on the pricing page.

Something wrong? Report an error with a public link that shows it; corrections re-run the review.

Top alternatives to Hy4 preview

See all
  • 67.1

    Qwen3.8 27B is an open-weight dense vision-language model from Qwen.

    ModelOpen source
  • 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

    ModelOpen source
  • 64.7

    Qwen3.8 Flash is a multimodal reasoning model from Alibaba.

    ModelOpen source
  • 64.3

    Qwen3.6 27B is a dense 27-billion-parameter language model from the Qwen Team at Alibaba, released in April 2026.

    ModelOpen source