Tencent Hy4 API: Price, Specs and How to Call Hunyuan's New Model (2026)

Short answer

TeamoRouter hasn't launched Hy4. If you want a Chinese model for coding and agent work with a TeamoRouter key today, use one of these:

• glm-5.3: Zhipu GLM-5.3 (model page)

• kimi-k3.1-pro: Moonshot AI Kimi K3.1 Pro (model page)

• deepseek-v4-pro: DeepSeek V4 Pro (model page)

• claude-sonnet-5-5, if you'd rather use Claude for the same agent work (model page)

About Hy4 itself: Tencent Hunyuan released the full version of Hy4 on November 4, 2026. Compared with Hy4 preview, which Tencent released with open weights on August 28, Tencent says the full version scales up pre-training and post-training, adds vision for the first time, and improves instruction following, long-document understanding, long-horizon tasks and deep research. Hy4 is live on Tencent Cloud and OpenRouter, and Tencent's API price is ¥6 per million input tokens, ¥0.3 cached input and ¥18 output, unchanged from Hy4 preview.

New accounts start with a $0 balance — top up before your first request. Top up from $5. If a request returns 402 (400 on the OpenAI format) with "wallet balance is insufficient", your balance is empty — your base URL and key are fine. Top up and retry.

Until Hy4 is on TeamoRouter, call GLM-5.3 or Kimi K3.1 Pro with the same key

Specs at a glance

From Tencent Hunyuan's release post, the Hy4 preview model card on Hugging Face and the OpenRouter listing, checked on November 4, 2026:

Price

These are the prices in Tencent Hunyuan's release post, which says the full version keeps Hy4 preview's API pricing. Tencent lists them in RMB; the Tencent Cloud TokenHub console shows what you are actually billed.

What Tencent says it's for

From Tencent Hunyuan's November 4 release post:

• Vision and computer use: Hy4 can combine image understanding, reasoning and tool calls for image processing and chart analysis. It is also the first Hy model that operates a computer, mixing the command line and the graphical interface, with targeted training on professional software such as Blender and CAD.
• Front-end and game development: support for engines such as UE, Unity and Godot. In Tencent's tests, Hy4 scored 19.1% higher on GameDev and 9.8% higher on GameLogic than Hy4 preview.
• Finance and office work: in a crowd test of 579 real finance tasks that Tencent ran with its fintech group, Tencent says Hy4 ranked first among open models.
• Limits: Tencent itself says Hy4 is still well behind the top closed models, especially on hard tests in frontier math, long-horizon development and visual question answering.

This section lists only what Tencent has published. The benchmarks and the crowd test were run by Tencent; we don't repeat third-party comparisons.

Safety notes and what's next

• Cybersecurity data: Tencent says it trained Hy4 only on defensive security data, such as PoC generation and vulnerability fixes, and removed offensive material such as exploitation, reverse engineering and penetration testing.
• Open weights: Tencent plans to release Hy4's weights. It is working with the vLLM and SGLang communities and Inferact on inference support and will publish a system card when the weights come out.
• Hy4 flash: Tencent plans to release Hy4 flash soon, aimed at lower cost and faster inference.

Calling Hy4 on Tencent Cloud or OpenRouter

Tencent's announcement lists two API routes:

• Tencent Cloud TokenHub: enable Hy4 in the Tencent Cloud console. Use the model name and base URL the console shows; you are billed at Tencent Cloud's prices.
• OpenRouter: the model ID is tencent/hy4, called with an OpenRouter key.

Hy4 supports tool calls. Before you connect it to a coding tool such as Claude Code or Codex, check that the route you use supports the API format that tool needs.

Use an alternative on TeamoRouter now

TeamoRouter's base URL is OpenAI-compatible, so the models above take the same request; only model changes:

curl https://api.teamorouter.com/v1/chat/completions \
  -H "Authorization: Bearer YOUR_TEAMOROUTER_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "glm-5.3",
    "messages": [{"role": "user", "content": "Find the bug in this function and suggest a fix: ..."}]
  }'

In coding agents, set the model name in the tool's config: Kimi K3.1 Pro with Claude Code, DeepSeek V4 Pro in Codex, Claude Code guide.

TeamoRouter bills per token with no subscription. See the pricing page for live prices.

Paying for usage

No subscription: you top up a balance from $5 and pay only for the tokens you use, by card or with USDT, USDC or ETH from any exchange app or wallet (how to pay with USDT). The same key works for every model in the catalog.

FAQ

Is Hy4 open source?

Not the full version yet. Tencent says it plans to release Hy4's weights. As of November 4, 2026, the Hy4 model you can download from Hugging Face is Hy4-preview, under Apache 2.0.

How much does Hy4 cost?

On Tencent's API it costs ¥6 per million input tokens, ¥0.3 cached input and ¥18 output, the same as Hy4 preview. The Tencent Cloud console shows the final billed price.

What changed from Hy4 preview?

Tencent says the full version was trained at a larger scale, adds image input and computer use, and is stronger at coding, game development, finance and long-document tasks, at the same API price.

Can I use Hy4 on TeamoRouter?

Not yet. glm-5.3, kimi-k3.1-pro and deepseek-v4-pro work today with the same key. Every live model is in the catalog.

Related

• Kimi K3.1 Pro API and Claude Code
• DeepSeek V4 Pro in Codex
• LongCat-2.5 API
• API integration guide