Tencent: Hy3

Tencent Hy reasoning model for coding, instruction following, and agent tasks

hy3

AvailableTencentChatNew

Input / Output: $0.132 / $0.528
Context: 197K
Released: Jul 5, 2026
Max output: 128K
Modalities: Chat
Capabilities: Tool usePrompt CacheReasoning

About Hy3

Hy3 is Tencent Hunyuan's open-weight reasoning model for coding, office work, and agent tasks. It is a mixture-of-experts design with 295B total and 21B active parameters, released under Apache 2.0 on July 6, 2026. It handles text only and offers three reasoning settings, so it suits agent backends that need a smaller alternative to Hy4 Preview.

Where it works well

Coding: Tencent's model card reports 78% on SWE-Bench Verified and 57.9% on SWE-Bench Pro.
Reasoning: the card lists 90.4 on GPQA Diamond.
Agent tasks: it reports 71.7% on Terminal Bench 2.1.
Reliability: Tencent says hallucinations and multi-turn issues each fell by more than half versus its preview.
Reasoning effort is selectable: no_think, low, or high.

When to choose another model

It is text-only; image, audio, and video inputs need another model.
Hy4 Preview is Tencent's newer and larger model, with a much longer context, for tasks that outgrow Hy3.
Self-hosting the full weights needs a multi-GPU server, so most developers call it through an API.

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": "hy3", "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

per 1M tokens

Official price: Input $0.132 / Output $0.528 / Cache read $0.033
Verified price: —
Discount: —

Prompt cache pricing

Cache read

Official price: $0.033
Verified price: —
Discount: —

	Official priceper 1M tokens	Verified priceper 1M tokens	Discount
Default	Input $0.132 / Output $0.528 / Cache read $0.033	—	—
Prompt cache pricing
Cache read	$0.033	—	—

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 1, 2026, 6:33 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 Hy3 in Console with a prompt ready to edit or send.

Help me try hy3 with a short message at /v1/responses. Show the reply, latency, and cost.

Use cases

Best for

Reasoning

Office and finance modeling
Build spreadsheets-as-code, draft financial models, and explain assumptions, a use Tencent highlights alongside coding.
Frontend generation
Generate component code and page layouts from a short design brief, then iterate on feedback.
Agent backends
Run tool-calling loops for ticket triage or data lookups, switching reasoning to no_think for simple steps.
Code review and bug fixing
Read a diff, find defects, and propose patches, with high reasoning for subtle bugs.

Prompt examples

Build a three-statement financial model skeleton for a SaaS company with monthly cohorts, and list the assumptions I need to fill in.

Review this pull request diff for race conditions and error-handling gaps, and rank the issues by severity.

Create a React pricing table component with a monthly and yearly toggle and accessible keyboard controls.

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 exactly is Hy3?

Hy3 is Tencent's Hunyuan model released on July 6, 2026 as an open-weight reasoning model. It uses a mixture-of-experts design with 295B total and 21B active parameters, and Tencent publishes it under Apache 2.0 on Hugging Face and ModelScope.

What reasoning modes does Hy3 offer?

Tencent's card lists reasoning effort values of no_think for direct answers, low, and high for increasingly deep chain-of-thought. Use no_think for lookups and formatting, and high for debugging, math, or planning.

Can Hy3 process images?

No. The model card describes a text-only model, and it takes no image input. For screenshots or documents as images, pick a vision-capable model. Reasoning and tool calling work on text content only.

Is Hy3 good at coding?

Tencent's model card reports 78% on SWE-Bench Verified and 57.9% on SWE-Bench Pro, and it names coding as a core strength. Those figures are the maker's own, so run your repositories through it before switching production agents.

Hy3 or Hy4 Preview?

Hy4 Preview is the larger, newer model with a far longer context and stronger reported engineering results, but it is still a preview. Hy3 is the stable release, which suits steadier workloads.

How much does Hy3 cost?

On TokenLab, Hy3 costs Input $0.132 / Output $0.528 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 Hy3?

Hy3 accepts up to 196,608 tokens of context and returns up to 131,072 tokens in one response.

Which endpoint should Hy3 use?

Use https://api.tokenlab.sh/v1/responses for Hy3. The request example below shows the matching code shape.

Which operations does Hy3 support?

Hy3 supports Text to text. Select an operation above to see its endpoint and request example.