hy4-preview

hy4-preview is the preview channel of the Hunyuan 4 line on the platform side: the catalog labels its context at 1,048,576 (1M), the longest window on the Hunyuan 4 line. It runs in parallel with the regular hy4 channel, and the difference is channel positioning and window labelling rather than a stripped-down official build.

Specification

Item	Value
Platform model ID	hy4-preview
Channel type	Preview channel, in parallel with hy4
Catalog context label	1,048,576 (1M)
Matching local weights	The Hunyuan-4B series under the tencent organisation
Local weight formats	BF16 / FP8 / GPTQ-Int4 / AWQ-Int4
Community GGUF	Multiple authors, from 2.61GB for Q4_K_M

How it differs from the regular hy4 channel

Dimension	hy4	hy4-preview
Channel positioning	Regular channel	Preview channel
Catalog context label	currently unlabeled	1,048,576 (1M)
Fits	Everyday chat, classification, rewriting	One-shot processing of very long material
Local weights	The same Hunyuan-4B series	The same Hunyuan-4B series

Three things to keep in mind

1. 1M is a ceiling, not a promise. Retrieval accuracy, attention dilution, and inference cost at long windows all need measuring on your own data.
2. Derive your local window from memory. The KV cache grows linearly with context length, and filling the window consumes tens of gigabytes, so the practical range is far below the label.
3. Do not mix the two naming systems. Platform channel names such as hy4-preview and HuggingFace weight names such as the Hunyuan-4B series are separate worlds.

Related reading

• hy4 model page — specification and quantization matrix for the regular channel
• hy4 model selection guide — choosing between channels and sizes
• hy4 GGUF local deployment — quantization tiers and memory math
• Tencent Hunyuan family overview — the full family
• VRAM calculation guide — the general memory math

from openai import OpenAI

client = OpenAI(
    api_key="YOUR_API_KEY",
    base_url=" echo exaiv.com;; *) echo https://exaiv.com;; esac)/api/ai/v1"
)

response = client.chat.completions.create(
    model="hy4-preview",
    messages=[
        {"role": "system", "content": "You are a helpful assistant."},
        {"role": "user", "content": "Hello, introduce yourself"}
    ]
)
print(response.choices[0].message.content)

stream = client.chat.completions.create(
    model="hy4-preview",
    messages=[{"role": "user", "content": "Write a short poem about autumn"}],
    stream=True,
)
for chunk in stream:
    print(chunk.choices[0].delta.content or "", end="")

curl  echo exaiv.com;; *) echo https://exaiv.com;; esac)/api/ai/v1/chat/completions \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "hy4-preview",
    "messages": [{"role": "user", "content": "Hello"}],
    "stream": false
  }'