Hy4-preview

Hy4 preview is a 770B-parameter Mixture-of-Experts flagship model from the Tencent Hy Team, with 49B parameters activated per token and a 1M-token context window. It uses Gated DeepSeek Sparse Attention with IndexCache, and is post-trained for software engineering, office and data analysis, game development, and scientific research.

Context
1.0M
Input price
$0.83/1M tokens
Output price
$2.5/1M tokens

Capabilities

Reasoning & Interaction
Reasoning
Tool Calling
Output Control
Structured Output
JSON Mode
Infrastructure
Streaming

Details

Model IDtencent/Hy4-preview
Provider DeepInfra
Creator tencent
Family—
License—
Statusactive
Typechat
Input modalitiestext
Output modalitiestext
Open weightYes
Parameters—
Architecture—
Context window1.0M tokens
Max input—
Max output—
Reasoning tokensYes
Knowledge cutoff—
Release date—
Sourceofficial

Tools

Function Callingfunction_calling
Call external functions and APIs

Endpoints

Chat CompletionsPOST
Generate chat responses with messageshttps://api.deepinfra.com/v1/openai/v1/chat/completions

Pricing

Input
$0.83
Output
$2.5
Cache write
—
Cache read
$0.04
Batch in
—
Batch out
—

API

GET/v1/models/deepinfra/tencent/Hy4-preview