Hy4

Tencent’s Hy4 preview, for long-horizon software work: understanding, planning, debugging and verifying. Also office and analysis work, with a context of a million tokens.

Get an API key API & docs

Hy4 Tencent

Context: 1M; tokens
Inputs: Text

Best for

Software engineering
Office and analysis
Research and games

What it does best.

Software engineering: Long development tasks, from understanding a codebase to verifying the change.
Office and analysis: Reports, documents and the analysis behind them.
Research and games: Scientific research and game development, the card’s other two uses.

It mines while it answers.

On the Peryn API, every request mines as much $PEARL as it can while it is served: inside the AI work, and in the gaps between requests.

How Peryn works

The model.

Made by: Tencent
Context: 1M tokens, up to 64K out
Size: 770B parameters, mixture of experts: 49B active per token
Inputs: Text
Licence: Apache 2.0
Official name: Hy4 preview

From Tencent’s official model card.

One call, OpenAI-compatible.

Use the OpenAI SDK: set our base URL and your key, and keep your code.

import osfrom openai import OpenAI client = OpenAI(base_url="https://api.peryn.ai/v1",api_key=os.environ["PERYN_API_KEY"],) stream = client.chat.completions.create(model="hy4-preview",messages=[{"role": "user", "content": "Hello"}],stream=True,)for chunk in stream:if chunk.choices:print(chunk.choices[0].delta.content or "", end="")

Other models on Peryn: DeepSeek V4.1 Flash · GLM-5.3-Flash · MiniMax M3.1 Flash