Table of Contents
Chutes has launched Tencent Hy4 as a Tencent Hunyuan launch partner, adding a flagship model that can interpret images and work through terminal and desktop tasks. Access costs $0.525 per million input tokens and $1.5625 per million output tokens, with a separate $0.025 rate for cached input.
The release gives developers using Chutes’ OpenAI-compatible API a way to bring visual understanding into applications that already call its language models. Existing customers can select tencent/Hy4 as the model name, using their Chutes account and API key.
Hy4 builds on Tencent’s earlier preview with expanded pre-training and post-training, vision, and stronger long-running task execution. Tencent released the model on November 4; Chutes’ integration brings it to another hosted endpoint two days later.
From Image Understanding to Computer Tasks
Hy4 combines visual perception, reasoning, and tool use to interpret fine image details and complex charts. That expands the information an agent can work with beyond text: screenshots, application interfaces, and visual data can become inputs to a task rather than material a user must describe manually.
Tencent has also added computer-use capabilities spanning command-line and graphical interfaces. The company trained Hy4 specifically on professional software, including Blender and CAD applications, and on extended workflows that cross application boundaries. These capabilities target work that involves planning several actions, using tools, and continuing toward a result across multiple steps.
In Tencent’s evaluations and development work, the model’s visual capabilities also support frontend engineering and game development. Training for frontend work covered interface interactions, business logic, state management, and full-stack integration. For games, Tencent worked with professional developers to build training data and evaluations around feature implementation, gameplay logic, state tracking, and asset pipelines, including work with Unreal Engine, Unity, and Godot.
Chutes Pricing and API Access
The Hy4 endpoint on Chutes lists three token rates:
| Token Category | Price per Million Tokens |
|---|---|
| Input | $0.525 |
| Output | $1.5625 |
| Cached Input | $0.025 |
Input tokens cover material sent to the model, while output tokens cover its generated response. The cached-input rate applies to input served from cache, such as when an application repeatedly supplies shared context. Actual request costs depend on the amounts of input, generated output, and cached input used.
Tencent lists its own API rates at $0.834 per million input tokens, $2.501 per million output tokens, and $0.042 for cached input. Chutes’ published rates are lower in all three categories.
Developers can point an OpenAI-compatible SDK at https://llm.chutes.ai/v1, supply a Chutes API key, and set the model to tencent/Hy4. The endpoint is POST /v1/chat/completions, with authentication through Authorization: Bearer $CHUTES_API_KEY.
That preserves the familiar chat-completions request structure for applications already using compatible clients. Developers adding image inputs or computer-interaction workflows also need to provide the relevant inputs and tools in their application; selecting the model determines which system processes those requests.
