LLM Service TokenHub
LLM Service TokenHub is committed to providing enterprises and developers with a unified gateway for large model services. It integrates Tencent's proprietary Hy large model capabilities and introduces premium third-party models, covering scenarios such as general conversation, deep reasoning, code generation, visual understanding, image generation, video generation, 3D generation, and voice generation. The platform supports pay-as-you-go, guaranteed resources, and dedicated deployment, helping enterprises access AI computing power simply and efficiently.
FEATURES
Discover the key features
Tencent Hy4 preview
A next-generation productivity model with significantly enhanced Agent and complex task execution capabilities.
Tencent Hy4 preview
A next-generation productivity model with significantly enhanced Agent and complex task execution capabilities.
GLM-5.3-FlashX
GLM-5.3-FlashX is a multimodal understanding model from Zhipu AI, featuring 1M context length and inference speeds of up to 200 tokens/s, delivering a faster and smoother model experience.
GLM-5.3-FlashX
GLM-5.3-FlashX is a multimodal understanding model from Zhipu AI, featuring 1M context length and inference speeds of up to 200 tokens/s, delivering a faster and smoother model experience.
GLM-5.3-Flash
Zhipu's first natively multimodal model, delivering intelligence that surpasses GLM-5.2 through an ultra-low-cost architecture.
GLM-5.3-Flash
Zhipu's first natively multimodal model, delivering intelligence that surpasses GLM-5.2 through an ultra-low-cost architecture.
DeepSeek-V4.1-Flash
DeepSeek’s latest multimodal model has comprehensively outperformed V4 Pro across metrics including performance, cost, speed, and total latency.
DeepSeek-V4.1-Flash
DeepSeek’s latest multimodal model has comprehensively outperformed V4 Pro across metrics including performance, cost, speed, and total latency.
DeepSeek-V4-Pro
1.6 trillion-parameter MoE flagship with native 1M-token context, purpose-built for complex workflows.
DeepSeek-V4-Pro
1.6 trillion-parameter MoE flagship with native 1M-token context, purpose-built for complex workflows.
MiniMax-Video-H3
Native Multimodal Understanding and Generation: Supports diverse inputs and outputs—including text, images, audio, and video—to enable integrated content creation.
MiniMax-Video-H3
Native Multimodal Understanding and Generation: Supports diverse inputs and outputs—including text, images, audio, and video—to enable integrated content creation.
Tencent Hy4 preview
A next-generation productivity model with significantly enhanced Agent and complex task execution capabilities.
GLM-5.3-FlashX
GLM-5.3-FlashX is a multimodal understanding model from Zhipu AI, featuring 1M context length and inference speeds of up to 200 tokens/s, delivering a faster and smoother model experience.
GLM-5.3-Flash
Zhipu's first natively multimodal model, delivering intelligence that surpasses GLM-5.2 through an ultra-low-cost architecture.
DeepSeek-V4.1-Flash
DeepSeek’s latest multimodal model has comprehensively outperformed V4 Pro across metrics including performance, cost, speed, and total latency.
DeepSeek-V4-Pro
1.6 trillion-parameter MoE flagship with native 1M-token context, purpose-built for complex workflows.
MiniMax-Video-H3
Native Multimodal Understanding and Generation: Supports diverse inputs and outputs—including text, images, audio, and video—to enable integrated content creation.
View All
BENEFITS
Why Choose TokenHub
Rich Models, Competitive Pricing
In addition to Tencent’s Hunyuan, YouTu, and other proprietary models, the platform introduces popular third-party models. It supports multiple purchasing modes to meet usage needs from business validation to large-scale production, helping you flexibly balance cost and stability.
Leading Multimodal Technology
Powered by advanced proprietary algorithms, the technology supports the generation of diverse digital content—including images, video, audio, text, and 3D assets. The generated content features high clarity and natural visuals that adhere to the laws of the physical world.
Seamless Ecosystem Integration
Integrates with Tencent’s large model training, agent infrastructure, knowledge engine, and agent building capabilities, lowering the barrier for developers and forming a complete closed loop from model to application.
Rich Models, Competitive Pricing
In addition to Tencent’s Hunyuan, YouTu, and other proprietary models, the platform introduces popular third-party models. It supports multiple purchasing modes to meet usage needs from business validation to large-scale production, helping you flexibly balance cost and stability.
Leading Multimodal Technology
Powered by advanced proprietary algorithms, the technology supports the generation of diverse digital content—including images, video, audio, text, and 3D assets. The generated content features high clarity and natural visuals that adhere to the laws of the physical world.
Seamless Ecosystem Integration
Integrates with Tencent’s large model training, agent infrastructure, knowledge engine, and agent building capabilities, lowering the barrier for developers and forming a complete closed loop from model to application.
LARGE LANGUAGE MODEL
Tencent Cloud Large Language Model Token Plan（Coming Soon）
Aggregates multiple mainstream models, compatible with various mainstream Claw and AI programming tools, empowering every line of your code with AI.
Mainstream Model Collection
GLM-5.2、Kimi-K2.6、DeepSeek-V4、MiniMax-M3 with more models continuously being onboarded.
Compatible with Claw and Coding Tools
Supports OpenClaw, Workbuddy, QClaw and other native Claw scenarios for office automation. Compatible with Codebuddy, Claude Code, Cursor and other mainstream AI programming tools, adapting to various development scenarios.
Same Usage, Better Price
Exceptional value for money — calling the same models via a plan is cheaper than pay-as-you-go API billing. Within the plan allowance, the more you use, the more you save.
LARGE LANGUAGE MODEL
Tencent Cloud Large Language Model Token Plan（Coming Soon）
Aggregates multiple mainstream models, compatible with various mainstream Claw and AI programming tools, empowering every line of your code with AI.
Mainstream Model Collection
GLM-5.2、Kimi-K2.6、DeepSeek-V4、MiniMax-M3 with more models continuously being onboarded.
Compatible with Claw and Coding Tools
Supports OpenClaw, Workbuddy, QClaw and other native Claw scenarios for office automation. Compatible with Codebuddy, Claude Code, Cursor and other mainstream AI programming tools, adapting to various development scenarios.
Same Usage, Better Price
Exceptional value for money — calling the same models via a plan is cheaper than pay-as-you-go API billing. Within the plan allowance, the more you use, the more you save.
GETTING STARTED
A Collection of Mainstream Programming Models
01
Hy Models
Hy4 preview，Hy3，Hy-Image-v3
02
Third-Party Models
DeepSeek、Kimi、GLM、MiniMax、Qwen
PRODUCTS
Variety of products for your needs
Text Generation
Through natural language conversation, it can combine images and model outputs to generate relevant content, widely used in scenarios such as text generation, creative writing, and Q&A. The platform offers Hunyuan and popular third-party models for you to choose from.
- Text Creation & Processing — Supports copywriting, rewriting & polishing, content summarization, multilingual translation, long-text analysis, and more.
- Dialogue & Q&A — Provides intelligent conversation, knowledge Q&A, logical reasoning, and solution recommendations.
- Code & Technical Generation — Supports code writing, SQL generation, script development, and technical documentation authoring. Title: Code Generation
Image Creation Large Model
An API technical service providing AI image generation and processing capabilities, integrating text-to-image, single-image-to-image, and multi-image-to-image capabilities. It deeply understands intent with precise addition, deletion, modification, and expansion, maintaining high consistency and stable output for characters and scenes, providing technical support for high-quality image content creation and content operations.
- Powered by the Hy large-scale model, it delivers robust Chinese semantic understanding and precise control over imagery. It accurately generates highly consistent characters, scenes, and storyboard assets, meeting the demands for professional-grade, high-precision image creation.
- Integrating capabilities from cutting-edge global commercial models, it offers exceptional stylistic diversity and artistic expression. It enables the rapid exploration of diverse visual styles during asset production, generating images that are both creative and market-ready.
Video Creation Large Model
Based on Tencent's video generation large model and a series of leading audio/video AI technologies, supports high-quality video content generation and processing. Helps professional video creators reduce production costs and discover creative ideas while enhancing the fun of video-based social entertainment. Widely applicable to short video platforms, film & TV production, advertising & marketing, social media, gaming, and more.
- It supports Hy as well as mainstream third-party video generation models such as MiniMax, Kling, and Vidu, delivering high-resolution, stylistically diverse, and controllable output, alongside capabilities like video enhancement and object removal.
- Tailored to the needs of sectors such as education, advertising and marketing, and short-form drama, it offers a wide range of applications—including document-to-video conversion, video stylization (re-rendering), text-to-short-drama generation, viral content replication, and viral video distribution. These features are launching soon; stay tuned.
Tencent HY 3D Global
Tencent HY 3D is powered by a highly efficient, proprietary generative AI foundation model, offering developers professional and efficient API services for 3D content generation. Its applications span various fields, including 3D printing, film and animation, product design, digital twins, and e-commerce marketing. The Tencent HY 3D API significantly lowers the technical barriers to high-fidelity 3D production, enhancing content creation efficiency and design flexibility, thereby empowering diverse industries to achieve agile and productive 3D workflows.
- 3D Model Generation — Supports generating 3D models from various inputs, such as text, images, and sketches.
- World Models — Supports generating editable and interactive 3D scene files from text and images.
SCENARIO
Variety of solutions for your needs
Document Scenarios
Provides capabilities including document creation, text polishing, text proofreading, spreadsheet formula generation, and chart creation, improving creative efficiency and enhancing the creative experience.
Meeting Scenarios
Supports intelligent pre-meeting agenda preparation, real-time in-meeting recording and Q&A responses, and automatic post-meeting summarization of minutes and to-do items. Leverages large model capabilities to improve efficiency across the entire meeting workflow.
Advertising Scenarios
Based on the large model’s understanding of content commercial value, mines users’ potential business interests, provides intelligent ad recommendations and creative generation, and improves content relevance rates and ad conversion performance.
Marketing Scenarios
Supports intelligent generation of brand marketing content, user profiling analysis, personalized recommendations, and marketing strategy optimization, helping enterprises achieve precision marketing and efficient customer acquisition.
Creation Scenarios
For content creators, provides AI creation capabilities including short video production, micro-drama generation, film & TV content derivation, and virtual character operations. Powered by large models, it enables “describe-to-generate”, significantly lowering professional production barriers.
CUSTOMER STORIES
Trusted by partners worldwide
Tencent Meeting
Tencent Meeting is an audio and video conferencing application under Tencent Cloud. In February 2024, it launched an AI Assistant powered by the Hunyuan large model to help users with pre-meeting preparation, in-meeting reminders and real-time Q&A responses, and post-meeting summarization of meeting minutes and to-do items, greatly improving meeting efficiency.
Tencent Ads
Tencent Ads leverages the Hunyuan large model for understanding the commercial value of content, mining users’ potential business interests, and providing recommendation insights in ad-based common-sense recommendations. It has achieved significant improvements in content relevance rates and A/B test ad GMV.
Trusted by partners worldwide
