

Use the Tencent API on Atlas Cloud to run Hunyuan 3 for reasoning and Hunyuan 3D for mesh generation. One key, one bill, pay as you go for both models today.
Tencent is developed by Tencent. Atlas Cloud (operated by Atlas Cloud AI LLC) provides access to it and does not own it. All trademarks belong to their respective owners.
Generate cinematic, high-fidelity videos from text and images with the latest AI video generation models on Atlas Cloud.

Tencent's WAND Vega Video 1.5 family is part of WAND VOD's international offering. Beyond reference inputs, text prompts can generate short clips, with pro and lite variants available across all three generation modes. Atlas Cloud provides access to these six models at standard base rates of $0.045 per output second for lite and $0.093 for pro. Start building today.

WAND Vega Video 1.0 is part of Tencent's WAND VOD offering for the international market. Turn prompts into clips, animate a first frame with an optional last frame, or use reference media to guide video generation. On Atlas Cloud, standard pricing starts at $0.032 per output second for Pro Mini, giving developers a clear starting point for estimating generation costs. Start building today.
Create stunning, production-ready visuals from prompts and references using state-of-the-art AI image generation models on Atlas Cloud.

The Hunyuan 3D API puts Tencent's Hunyuan3D foundation model in your stack, turning a text prompt or a single image into clean, watertight geometry through a two-stage shape and texture pipeline. Pick the Pro tier for up to 1.5M faces and 4K PBR materials, or Rapid for 2 to 3 minute iteration. Atlas Cloud serves both tiers through one OpenAI-compatible key with transparent pay-as-you-go pricing from $0.02 per generation and Day-0 access. Start building today.

WAND Vega Image 1.5 is Tencent Cloud's image generation and editing family for the WAND VOD international site. Turn text prompts into images, or guide edits with text instructions and optional reference images. On Atlas Cloud, both models carry a standard base price of $0.067 per image, giving you a clear starting point for budgeting generation and editing. Start building today.

Tencent Cloud's WAND Vega Image 1.0 belongs to the WAND VOD image family available on the international site. Guide edits of existing visuals using text instructions and optional reference images. Atlas Cloud brings generation and editing models together, with standard base pricing starting at $0.022 per image for Lite to help developers budget for image production. Start building today.
Power chat, reasoning, and agents at scale with leading large language models, served fast and affordably on Atlas Cloud.

Tencent Hy4 is Tencent's next-generation Hunyuan model. It focuses on frontier reasoning, advanced agent capabilities, and complex task execution, supporting developers building agents for demanding workflows. On Atlas Cloud, input, output, and cache usage have separate rates, helping you plan API costs around the way your application uses the model. Start building today.

Tencent Hunyuan-Hy3 is an open-source Mixture-of-Experts (MoE) model featuring a integrated fast-and-slow thinking mechanism. With 295B total (21B active) parameters and a 256K context window, it delivers flagship-level performance with exceptional cost-effectiveness. Hy3 excels in complex Agent workflows, instruction following, and hallucination reduction, and is currently powering core Tencent services (like Tencent Yuanbao and Tencent Cloud) via low-cost APIs.
Compare standard vs. our pricing across every Tencent model.
| Model | Standard Price (USD) | Our Price (USD) | Discount | |
|---|---|---|---|---|
| Hy4 | $0.834/$2.501per 1M tokens1048.6K context | $0.417/$1.25M in/outper 1M tokens1048.6K context | -50% | View |
| Hunyuan 3D Rapid Image-to-3D | $0.02/pic | Start from$0.02/pic | — | View |
| Hunyuan 3D Rapid Text-to-3D | $0.02/pic | Start from$0.02/pic | — | View |
| Hunyuan 3D Pro Image-to-3D | $0.02/pic | Start from$0.02/pic | — | View |
| Hunyuan 3D Pro Text-to-3D | $0.02/pic | Start from$0.02/pic | — | View |
| Tencent WAND Vega Video 1.5-pro Reference To Video | $0.093 | Start from$0.093/s video | — | View |
Instantly explore and experiment with production-ready models in the Atlas Playground. Start customizing with one click.
Teams use the Tencent API on Atlas Cloud to run agentic coding, document review, and 3D asset generation from a single account. Hy3 handles reasoning-heavy tasks while Hunyuan 3D turns text and images into production-ready meshes.
Hy3 accepts a full codebase inside its 256K context window, enough to reason across many files at once. Developers set reasoning_effort to high so it can trace a bug through the repo and propose a fix. On Atlas Cloud this runs through the same Tencent API endpoint used for chat, with no separate agent infrastructure to stand up.
Hy3 was trained to reduce fabricated clauses and misquotes, which matters when the input is a long contract or regulatory filing. The 256K context window lets it read the full document instead of chunking it into fragments. Legal and compliance teams use this to flag risky clauses before a human reviews the file.
Studios feed in a character sketch or reference image and get back a mesh with up to 1.5M faces and 4K PBR textures. Artists then refine the output instead of modeling from a blank scene. Unity China and other studios already use Hunyuan 3D this way to cut early production time.
Retailers convert standard product photos into interactive 3D views customers can rotate and zoom. In a furniture trial, Tencent reported a 35 percent lift in click-through rate for listings with a 3D model attached. The GLB and USDZ exports plug directly into most storefront and AR viewers.
Makers and 3D printing teams use the Rapid tier to turn a sketch or photo into a printable mesh in 2 to 3 minutes. Each job includes an MP4 preview so you can check the shape before committing to a full-resolution export. Bambu Lab is among the companies building this kind of workflow on top of Hunyuan 3D.
Hy3 was tuned for tool calling and task planning, not just chat responses, with fixed reliability issues around output formatting. It already drives Tencent's own OS-level assistant for tasks like file editing, file management, and diagnostics. Developers can plug the same Tencent API into their own agent scaffolding on Atlas Cloud.
Hy3 API is Tencent's flagship reasoning model, built for complex problem solving and agentic workflows. It runs on a 295 billion parameter mixture-of-experts architecture with 21 billion active parameters and a 256K context window. On Atlas Cloud you call it through an OpenAI-compatible Tencent API endpoint alongside 400+ other models.
Yes. Hy3 was tuned specifically for agents, not just question answering, and it already powers Tencent's own CodeBuddy assistant in production. Developers can set reasoning_effort to high for multi-step coding tasks or no_think for fast direct answers.
Yes. Tencent open-sourced the final Hy3 release under the Apache 2.0 license after collecting feedback from more than 50 product teams during its preview. Atlas Cloud hosts the Hy3 API so you can call it directly instead of running the weights yourself.
Hunyuan 3D generations start at $0.02 each, billed pay-as-you-go with no subscription required. Hy3 API usage is billed the same way, per token, through the same Atlas Cloud account. One API key covers both models and every other model on the platform.
Guides, tutorials, and product updates to help you get the most out of Atlas Cloud.