tencent cloud

One API. Every flagship model
Token Plan Hy4 at 75% off

Subscribe to one plan and access Hunyuan, DeepSeek, GLM, Kimi & more — unified billing, plug and play.

  • Tier-1 performance — ultra-low latency and fast TTFT
  • OpenAI-compatible — standard /v1/chat/completions & SSE streaming
  • DeepSeek official integration — direct from the source
Hunyuan DeepSeek GLM Kimi + more

Exclusive for new users · Up to 1 million free tokens, valid for 90 days

TokenHub multiple model calls

Pick a Plan, Get Started

Monthly subscription with pooled credits. Switch between models freely. Upgrade anytime.

Lite
Perfect for beginners
1,000 Credits/Month
Base quota 1×
$7 /Month
  • Ideal for first-time AI experience
  • Approximately 70 Q&A rounds
Subscribe Now
Pro
For developers and power users
7,900 Credits/Month
Approx. 7.9× quota
$51 /Month
  • For developers and productivity enthusiasts using AI daily
  • Approximately 600 Q&A rounds
Subscribe Now
Max
Heavy AI development
15,900 Credits/Month
Approx. 15.9× quota
$103 /Month
  • For power users who rely on AI as core productivity tool
  • Approximately 1,200 Q&A rounds
Subscribe Now

Top flagship models, switch freely

Call any model within the same plan, or pay per token. Transparent input/output dual pricing, synced with official docs.

Tencent Hy3
Coding & Agentic · Cost-leading
  • 295B total / 21B active MoE architecture
  • Native 256K context, balanced speed and cost
  • Enhanced coding, long-context, and multi-turn agent execution
$0.132/ 1M input
$0.528/ 1M output
Get Started
DeepSeek-V4-Pro
Flagship Agent · Deep reasoning
  • 1.6T total / 49B active MoE flagship architecture
  • 1M context + 384K output, three reasoning modes
  • Native Responses API, open-source SOTA on SWE-bench
$0.435/ 1M input
$0.87/ 1M output
Get Started
GLM-5.3
Coding SOTA · Cybersecurity
  • 743B MoE base + extreme post-training scaling
  • Open-source coding SOTA: 28.3 on Terminal Bench
  • Emergent cybersecurity: 84.5% on CyberGym
$1.4/ 1M input
$4.4/ 1M output
Get Started
Kimi K3.1 Pro
Programming & Agentic Flagship
  • 1M token context, native multimodal
  • Long-horizon programming & reasoning, excels at complex agent tasks
  • Supports 3D creation, covering end-to-end knowledge work
$2/ 1M input
$10/ 1M output
Get Started

Full model pricing

Singapore region. Promo pricing synced with official docs.

Model Type Input ($/1M) Output ($/1M) Cache hit ($/1M)
Tencent Hunyuan Hy4 New Agentic · Coding · Productivity $0.834 $2.501 $0.042 Activate
Tencent Hunyuan Hy3 General · Coding $0.132 $0.528 $0.033 Activate
DeepSeek-V4-Flash (0731) Cost-effective · High concurrency $0.14 $0.28 $0.0028 Activate
DeepSeek-V4-Pro (0813) Deep reasoning · Flagship $0.435 $0.87 $0.00363 Activate
DeepSeek-V3.2 Long-context SOTA $0.57 $1.71 $0.114 Activate
GLM-5.3 Coding SOTA $1.4 $4.4 $0.26 Activate
Kimi K3.1 Pro HighSpeed Ultra-fast inference · High concurrency $4 $20 $0.2 Activate
Kimi K3.1 Pro Flagship reasoning SOTA $2 $10 $0.2 Activate
Kimi K3 Engineering & reasoning SOTA $3 $15 $0.3 Activate
Kimi K2.7 Code Code-specialized $0.95 $4 $0.19 Activate
MiniMax-M3 Multimodal long-context $0.3 $1.2 $0.06 Activate
MiMo-V2.5-Pro Reasoning flagship $0.435 $0.87 $0.003 Activate

Since 2026/8/17, DeepSeek-V4 series uses peak/off-peak pricing; peak hours are 9:00-12:00 and 14:00-18:00 (Beijing time). GLM-5.3 launched 8/14 as open-source #1 in both coding and cybersecurity. Hy4 is now generally available, with multimodal understanding and agentic capabilities. See full pricing docs

All Your AI Tools Can Connect to TokenHub

In tools like CodeBuddy Code, WorkBuddy, Cursor, etc., simply enter TokenHub's API to call your plan and access models.

Agent Tools
OpenClaw
WorkBuddy
Hermes Agent
Programming IDE
CodeBuddy Code Recommended
Claude Code
OpenCode
Cline
Cursor
Kilo Code
Roo Code

What You Can Build

Text Generation
Image Generation
Video Generation
3D Generation
Intelligent Creation Agent

Text Generation

Through natural language dialogue, combined with images and model outputs, widely used in text generation, creation, Q&A and other scenarios. The platform provides Hunyuan and popular third-party models for you to choose from.

Text Creation & Processing

Supports copywriting, rewriting and polishing, content summarization, multilingual translation, long text analysis and other scenarios.

Dialogue & Q&A

Provides intelligent dialogue, knowledge Q&A, logical reasoning, solution recommendations and other capabilities.

Code & Technical Generation

Supports code writing, SQL generation, script development, and technical documentation writing.

Text Generation

Image Generation

Provides API technical services for AI image generation and processing, integrating text-to-image, single-image-to-image, multi-image-to-image and other capabilities. Deeply understands intent, precisely adds, deletes, modifies and expands, maintaining high consistency in characters and scenes for stable output, providing technical support for high-quality image content creation and content operations.

Hunyuan Image Generation (3.0)

Based on the Hunyuan large model, capable of thinking about image layout, composition, and brushstrokes, using world knowledge to reason about scenes that conform to common sense.

Hunyuan Image Generation (Turbo)

Based on the Hunyuan large model, adopts ultra-high compression codec to achieve fast response and high-quality image generation.

Image Generation

Video Generation

Based on Tencent Video Generation Large Model and a series of leading audio-visual AI technologies, supports high-quality video content generation or processing. It can help professional video creators reduce production costs and discover creative ideas, while enhancing the fun of video social entertainment. Widely applicable to short video platforms, film and television production, advertising marketing, social media, gaming and other fields.

Video Content Generation

Supports text or image input to generate HD video, achieving scene switching and multi-character interaction.

Portrait & Effect Driving

Single photo drives dynamic portrait video, supporting multi-style effect conversion.

AI Real Person Drama Generation

One-stop AI real person drama production, covering the full process of material generation, character driving, and content production.

Video Generation

3D Generation

Tencent Hunyuan 3D Generation (Tencent HY 3D) is based on self-developed generative AI large models, providing developers with professional and efficient 3D content generation API services. Can be applied to 3D printing, film and animation, product design, digital twins, e-commerce marketing and other fields. Tencent Hunyuan 3D API greatly reduces the technical threshold for high-quality 3D content production, improves creative efficiency and design freedom, and helps multiple industries achieve efficient and flexible 3D content production.

3D Model Generation

Supports text, images, sketches and other methods to generate 3D models.

Texture & Topology Optimization

Provides model optimization functions such as texture generation, intelligent topology, and UV unfolding.

Intelligent Component Generation

Automatically identifies model structure and generates corresponding 3D components.

3D Generation

Intelligent Creation Agent

Relying on core technologies such as Tencent Hunyuan large model, computer vision, and speech synthesis, with "hyper-realistic portrait restoration + full-process intelligent generation" as the core, creating a one-stop AI real person drama production system covering "material generation - character driving - content production".

Real Person Drama Generation

One-stop AI real person drama production, covering the full process of material generation, character driving, and content production.

Video Effect Generation

Upload images and select templates to quickly generate effect videos.

Intelligent Creation Agent

Before you buy, you may want to know

What is TokenHub?
TokenHub is Tencent Cloud's one-stop large model gateway. Through a single OpenAI-compatible API, one key, and one bill, you can access mainstream large models such as Hunyuan, DeepSeek, GLM, Kimi, and MiniMax. It does not change the models themselves, but provides a unified and stable entry point for calling them.
How can new users claim 1 million free Tokens?
Two ways: ① Log in to the TokenHub console, go to the Model Gallery and click "New User Free Trial" in the upper right corner, check the models you want to use and the credits will be credited; ② Or directly initiate an API call, the first call will automatically claim the corresponding model quota. Each main account can claim each model only once, and the event lasts until December 31, 2026.
Which models are covered by the free quota? How long is it valid?
Covers almost all mainstream models including DeepSeek-V4-Pro/Flash, GLM-5.2/5, Kimi-K2.6, MiniMax-M3, Hy-MT2, etc. Under the same main account, all models share a quota pool, with 1 million Tokens per model, valid for 90 days from the day of claim, expiring automatically if unused.
How long can 1 million Tokens last?
Estimated at approximately 4 characters per Token, 1 million Tokens equals about 4 million characters of input + output, enough to run through a prototype, tune your Prompt, and complete model selection validation. Code generation and long-text reasoning will consume faster, while lightweight conversations will last longer.
Will service be interrupted after the free quota is used up?
It depends on whether post-payment is enabled: if not enabled, service will automatically stop when the quota is exhausted; if post-payment is enabled in advance on the "Online Inference" page in the console, it will seamlessly switch to Token-based billing without interruption. It is recommended to enable it before the quota runs out.
Can one API switch between different models?
Yes. TokenHub is an OpenAI-compatible interface. Existing OpenAI client code only needs to change the base URL and key to access it. Switching models only requires changing one model name parameter, without needing to register separate accounts for each model or maintain multiple SDKs.

Start now — one plan, all flagship models.

New users get up to 1 million free tokens, validate your business at zero cost.