Tencent: Hy4

Hy4 is the general release of Tencent Hunyuan's flagship model and follows Hy4 Preview. It adds image understanding for charts, documents and multi-image questions, and it is built for agent work that calls tools and operates command-line and graphical interfaces. It reasons before it answers.

Compare models

hy4

AvailableTencentChatNew

Input / Output: $0.417 / $1.25
Context: 960K
Released: Nov 4, 2026
Max output: 64K
Modalities: VisionChat
Capabilities: Tool usePrompt CacheReasoning

About Hy4

Hy4 is Tencent Hunyuan's flagship model in its general release, following Hy4 Preview. Tencent released it on November 4, 2026. It adds visual understanding for the first time, with fine-grained image processing, complex chart analysis, and mathematical visual reasoning across documents and multiple images. It also adds computer use for command-line and graphical interfaces, alongside tool calling and reasoning.

Where it works well

Visual understanding is new in Hy4: Tencent describes fine-grained image processing, complex chart analysis, and mathematical visual reasoning.
It handles documents and several images in one task, which suits report pages, slide decks, and before-and-after comparisons.
Computer use for command-line and graphical interfaces lets an agent work in a terminal or from what is on screen.
Tool calling and reasoning are supported, and Tencent reports lower hallucination than Hy4 Preview.

When to choose another model

Visual understanding and computer use debut in this release, so test them on your own screens and documents before an agent runs unattended.
Tencent has announced the weights and a system card for later, so documentation beyond its launch materials is limited today.
Tencent plans a cheaper Hy4 Flash, which may suit cost-sensitive, high-volume work better once it is released.

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", "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 $0.417 / Output $1.25 / Cache read $0.021
Verified price: —
Discount: —

Prompt cache pricing

Cache read

Official price: $0.021
Verified price: —
Discount: —

	Official priceper 1M tokens	Verified priceper 1M tokens	Discount
Default spec	Input $0.417 / Output $1.25 / Cache read $0.021	—	—
Prompt cache pricing
Cache read	$0.021	—	—

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 5, 2026, 11:05 PM 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 in Console with a prompt ready to edit or send.

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

Use cases

Best for

Reasoning
Vision

Chart and report reading
Ask questions about a complex chart, a scanned report page, or several pages at once, and get answers tied to what the images show.
Math problems from diagrams
Work through geometry or physics problems given as pictures, using visual reasoning together with step-by-step reasoning.
Command-line agents
Let an agent run shell commands, read their output, and fix a failing build or script through a series of tool calls.
Screen-driven task automation
Have an agent read a screenshot of an application and decide the next entry or click in a repetitive workflow, such as filling a form.

Prompt examples

Here are two charts from our quarterly deck. Compare the trends, point out where they disagree, and list what to check in the source data.

This photo shows a geometry problem with a labeled triangle. Solve it step by step and state the answer last.

Run the test suite, read the failing output, and fix the first failure. Explain the change in two sentences.

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 the Hy4 model?

The general release of Tencent Hunyuan's flagship model, released on November 4, 2026 as the successor to Hy4 Preview. It adds image understanding and computer use, alongside tool calling and reasoning.

How does Hy4 differ from Hy4 Preview?

Hy4 adds visual understanding, covering fine-grained image processing, complex chart analysis, mathematical visual reasoning, documents, and multiple images. It also adds computer use for command-line and graphical interfaces, and Tencent reports lower hallucination than the preview. Hy4 Preview accepts text only.

What does computer use mean for Hy4?

Tencent says Hy4 can operate command-line and graphical interfaces, so an agent can drive a terminal or work from what is on screen. This is a capability of the model: your application still supplies the tools and carries out the actions the model requests.

Are the Hy4 weights available?

Not yet. Tencent has promised the weights and a system card for a later date. Until they appear, use Hy4 through an API, and check Tencent's own pages before you plan to self-host.

Is there a cheaper Hy4?

Tencent has announced a Hy4 Flash that it describes as cheaper than Hy4. It is not part of this release, so Hy4 is the model available for now, and Hy4 Preview remains a separate entry.

How much does Hy4 cost?

On TokenLab, Hy4 costs Input $0.417 / Output $1.25 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?

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

Which endpoint should Hy4 use?

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

Which operations does Hy4 support?

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