Tencent: Hy4 Preview

A next-generation productivity model with significantly enhanced Agent and complex task execution capabilities.

hy4-preview

AvailableTencentChatPreview

Input / Output: $0.834 / $2.50
Context: 960K
Max output: 64K
Modalities: Chat
Capabilities: Tool usePrompt CacheReasoning

About Hy4 Preview

Hy4 Preview is Tencent's early release of its next flagship open-weight language model, a 770B mixture-of-experts system with 49B active parameters, published under Apache 2.0 in late August 2026. It targets software engineering, office productivity, and scientific research with a very long context window. It handles text only and has two reasoning modes.

Where it works well

Software engineering: Tencent's model card reports 65.7% on SWE Bench Pro and 64.3% on Deep SWE.
Science: the card lists 92.3% on GPQA Diamond.
Two modes: high for chain-of-thought on complex work, and no_think for direct answers.
Tool calling and prompt caching are supported.

When to choose another model

Tencent calls it an early version and admits over-verification and weak reasoning on some complex tasks.
It is text-only; screenshots and documents as images need a vision model.
As a preview, behavior can change before the final Hy4 release.

Getting started

Create API key
Create a key in Console, then use it with every model on the platform.
02
Send your first request
Copy the example for your language and run it against the endpoint.
Text to textResponses API
POST/v1/responses
API format:
curl https://api.tokenlab.sh/v1/responses \ -H "Content-Type: application/json" \ -H "Authorization: Bearer sk-xxx" \ -d '{ "model": "hy4-preview", "input": "Hello!" }'

Pricing

The Verified price applies to Verified, which costs less on most models. The Official price is the model maker's published price and applies to the more reliable Official route. Auto bills the route that completes the request.

Default spec

per 1M tokens

Official price: Input $0.834 / Output $2.50 / Cache read $0.042
Verified price: —
Discount: —

Prompt cache pricing

Cache read

Official price: $0.042
Verified price: —
Discount: —

	Official priceper 1M tokens	Verified priceper 1M tokens	Discount
Default spec	Input $0.834 / Output $2.50 / Cache read $0.042	—	—
Prompt cache pricing
Cache read	$0.042	—	—

Usage & activity

Success rate is the share of requests that completed. Latency is how long a full response takes; P95 means 95% of requests finished within that time.

Usage & availability

Last 24 hours

Requests
Success rate
P95 latency
Total tokens

Model performance

Last updated Nov 13, 2026, 4:51 AM UTC

Metrics appear once privacy and data volume thresholds are met.

Data is based on aggregate user requests, excluding status checks.

Open in Console

Open Hy4 Preview in Console with a prompt ready to edit or send.

Help me try hy4-preview with a short message at /v1/responses. Show the reply, latency, and cost.

Use cases

Best for

Reasoning

Large-repository engineering
Hand an agent a big codebase and a long issue thread, and let it plan and apply changes across many files.
Scientific literature work
Read several papers in one prompt, compare methods, and draft a summary of open questions.
Office automation agents
Chain tool calls to fill reports, reconcile tables, and prepare slide outlines from raw notes.
Game development support
Prototype gameplay logic, tool scripts, and level-data generators, a use Tencent names for the model.

Prompt examples

Given the issue thread and the relevant modules, plan the change, list files to edit in order, and write the patch for the first one.

Compare the training setups described in these four papers and tell me which reported improvements are confounded.

Write a Python tool that validates level JSON files against our schema and reports every broken reference.

This model has conditional pricing. Monthly token totals alone cannot produce a reliable estimate; use the detailed pricing for the request specification, cache, and applicable time window.

FAQ

What is Hy4 Preview?

It is the early release of Tencent's Hy4 line, an open-weight mixture-of-experts model with 770B total and 49B active parameters. Tencent describes it as a preview with real headroom left in pre-training and post-training.

What reasoning modes does it have?

Tencent documents high, the default chain-of-thought mode for complex tasks, and no_think, which gives direct responses. Pick no_think for routine steps and keep high for planning, debugging, and research.

Does Hy4 Preview support images?

No. Tencent's card describes it as text generation only, and it takes no image input. Use a multimodal model for image or video input. Tool calling and prompt caching apply to text conversations.

How does it differ from Hy3?

Hy4 Preview is larger, accepts a much longer context, and reports higher engineering scores, but it is a preview. Hy3 is the stable release with a smaller footprint.

What are its known weaknesses?

Tencent names over-verification, where the model checks its work more than needed, and suboptimal reasoning on some complex tasks. Evaluate it on your own tasks before moving production workloads.

How much does Hy4 Preview cost?

On TokenLab, Hy4 Preview costs Input $0.834 / Output $2.50 per 1M tokens. The pricing table above shows the full breakdown. Rates depend on the billing unit, specification, and usage. Compare matching conditions in the model's detailed pricing; a single rate does not determine the total cost.

What are the context window and output limit of Hy4 Preview?

Hy4 Preview accepts up to 960,000 tokens of context and returns up to 64,000 tokens in one response.

Which endpoint should Hy4 Preview use?

Use https://api.tokenlab.sh/v1/responses for Hy4 Preview. The request example below shows the matching code shape.

Which operations does Hy4 Preview support?

Hy4 Preview supports Text to text. Select an operation above to see its endpoint and request example.