What model ID should I use for Tencent Hy3 Model API calls?

Tencent Cloud TokenHub documentation lists hy3 as the model ID for Tencent Hy3 Model. Recheck the current TokenHub documentation before production use.

Can Tencent Hy3 Model run locally?

Yes, Tencent Hy3 Model can be served locally from the published model resources, but it is a large 295B-parameter MoE model and requires serious infrastructure planning.

Step-by-step guide

How to Use Tencent Hy3 Model

Q: How do I use Tencent Hy3 Model?

Use Tencent Hy3 Model through the official research page and repository for model information, Tencent Cloud TokenHub for managed API access, or the official model hubs and serving frameworks for local deployment.

Tencent Hy3 Model can be used in three practical ways: read the official research and repository material, call the managed Tencent Cloud TokenHub API with model ID hy3, or run the open weights with a serving stack such as vLLM or SGLang.

Start with the quick answer Use the API Browse all articles

Fact check: July 16, 2026. Pricing, account eligibility, regions, rate limits, model IDs, and legal terms can change, so verify them in Tencent Cloud before production use.

Independent note: This page is not Tencent, Tencent Cloud, or an authorized support channel. Use official Tencent pages for accounts, API keys, billing, downloads, licenses, and incidents.

Quick Answer

How do you use Tencent Hy3 Model?

To use Tencent Hy3 Model, start from the official Tencent Hy3 Model research page and the Tencent-Hunyuan/Hy3 repository, then choose one of two implementation paths. For managed inference, follow Tencent Cloud TokenHub API setup and call model ID hy3. For private or local serving, download the published model weights from Hugging Face or ModelScope and deploy with supported serving tools.

There is no need to guess from random mirrors. Keep the access path direct: official page, official repository, official model hubs, TokenHub documentation, then your own evaluation and safety checks.

Use path at a glance

Managed API

Best when you want a faster prototype, server-side API key handling, usage billing, and less model-serving infrastructure.

Local or private serving

Best when you need open-weight control, private infrastructure, custom serving settings, or deeper deployment review.

Step 1

Choose the right Tencent Hy3 Model access path

Learn and evaluate: read the official research page, repository, model cards, and this guide before you build anything user-facing.
Prototype through API: use Tencent Cloud TokenHub when you want OpenAI-compatible requests and managed inference.
Self-host: use the open weights when deployment control matters enough to justify GPU planning, serving operations, monitoring, and safety review.

Step 2

Start from official Tencent Hy3 Model sources

Use the official Tencent Hy3 Model page for release context, the Tencent-Hunyuan/Hy3 repository for implementation notes, and official model pages on Hugging Face or ModelScope for weights. Avoid third-party pages that ask for keys, repack downloads, or claim special access without a clear source trail.

Use official Tencent and Tencent Cloud pages for account, billing, API key, privacy, and support decisions: start with Tencent Hy3 Model, TokenHub API setup, and TokenHub pricing.
Use repository-linked model hubs for downloads and model card checks: Tencent Hy3 Model on Hugging Face and Tencent Hy3 Model on ModelScope.
Record the model ID, weight source, license, and date of verification for production review.

Step 3

Use Tencent Hy3 Model through TokenHub API

Tencent Cloud TokenHub documents OpenAI-compatible access for language models. For Tencent Hy3 Model, the practical starting point is the model ID hy3, base URL https://tokenhub.tencentmaas.com/v1, and a Bearer API key from Tencent Cloud.

from openai import OpenAI

client = OpenAI(
    api_key="YOUR_TOKENHUB_API_KEY",
    base_url="https://tokenhub.tencentmaas.com/v1",
)

response = client.chat.completions.create(
    model="hy3",
    messages=[
        {"role": "system", "content": "You are a concise technical assistant."},
        {"role": "user", "content": "Explain Tencent Hy3 Model in three bullets."},
    ],
)

print(response.choices[0].message.content)

Keep API keys server-side, set budget alerts early, log request IDs, and test your own prompts before assuming the model is production-ready for your domain.

Step 4

Run Tencent Hy3 Model locally or privately

Local serving is useful when you need more control over data flow, deployment region, latency, customization, or cost at scale. It is not the easiest starting point. Tencent Hy3 Model is a large MoE model, so plan hardware, interconnect, quantization, context length, throughput, and observability before you promise a local deployment.

Start from the official repository and linked model hubs, not repackaged downloads.
Compare BF16 and FP8 options against your quality, memory, and latency targets.
Use serving frameworks such as vLLM or SGLang when they match the repository guidance.
Run evaluations for long context, tool calls, structured output, multilingual prompts, and your domain data.

Step 5

Write your first Tencent Hy3 Model prompt

Good prompts are specific about the job, evidence, constraints, output format, and uncertainty behavior. Do not rely on magic phrases. Treat the model as a capable component inside a system that still needs retrieval, validation, logging, and human review for high-impact work.

Task: Summarize the attached release notes for an engineering manager.

Constraints:
- Use five bullets or fewer.
- Separate confirmed facts from risks.
- Include open questions at the end.
- If the source does not answer something, say "not stated".

Step 6

Use Tencent Hy3 Model safely in production

Before a production rollout, decide what data can enter prompts, who can use the feature, how much it can spend, which tools the model can call, and which answers require human approval. Model quality matters, but the surrounding system usually decides whether the deployment is trustworthy.

Never paste secrets, passwords, private keys, unreleased financials, or customer data without explicit approval.
Validate structured outputs with schemas instead of trusting raw model text.
Use allowlists and approvals for tool calls, file writes, payments, messages, and account changes.
Track latency, error rate, token volume, cache hits, cost, and tool-call failures.

FAQ

How to use Tencent Hy3 Model: common questions

How do I use Tencent Hy3 Model?

Use the official Tencent Hy3 Model page and repository for source material, TokenHub API for managed inference, or official model hubs such as Hugging Face and ModelScope for local deployment.

What is the Tencent Hy3 Model API model ID?

Tencent Cloud TokenHub documentation lists hy3 as the model ID. Recheck current TokenHub documentation before production use.

Can I use Tencent Hy3 Model without an API?

Yes, if you use the published open weights and have suitable infrastructure. For many teams, the managed API is the faster first test.

Is there a Tencent Hy3 Model chat app?

This guide does not confirm a dedicated consumer chat app. Use official Tencent pages to verify any product surface before signing in or entering data.

What should I build first with Tencent Hy3 Model?

Start with a low-risk workflow such as summarization, internal Q&A, classification, extraction, or code-review assistance, then add evaluation and monitoring before expanding.

Next Reads

Go deeper after the first setup