Back to all plans
混元 API
Hy3 preview
TokenHub open API with Hy3 preview flagship plus Hunyuan vision, image, video, and 3D stack
Token API
SubscriptionToken API
Core models
Hy3 previewHy-MT2-ProHy-MT2-PlusHy-MT2-LiteHunyuan-roleHY-Image-V3.0HY-Image-LiteHY-Video-1.5YT-Video-2.0YT-Video-FXYT-Video-HumanActorHY-3D-3.0HY-3D-3.1HY-3D-ExpressYT-VITAHY-Vision-2.0-InstructHY-Vision-1.5-ThinkingHY-Vision-Video
Hy3 preview
TokenHub flagship language model—≤16K input ¥1.2/¥4 per 1M tokens with 16K–32K and 32K+ bands and ¥0.4 cache hits.
Hy-MT2-Pro
Hunyuan translation Pro—¥0.5 input, ¥2 output per 1M tokens.
Hy-MT2-Plus
Hunyuan translation Plus—¥0.5/¥2 per 1M tokens, same unit rates as Pro.
Hy-MT2-Lite
Hunyuan translation Lite—¥0.3/¥1.2 per 1M tokens, low-cost translation entry.
Hunyuan-role
Character roleplay vertical model—¥2.4/¥9.6 per 1M tokens.
See the official site for more models
Additional core model names still appear above, with full details on the latest official page.
Plan details
Hy3 preview
Language flagshipRecommendedInput
¥1.2
Output
¥4
Usage
Hy3 preview is the current TokenHub language workhorse—≤16K band ¥1.2/¥4 per 1M for agents, code, and long context.
Models
Longer inputs move to higher bands (16K–32K, 32K+)—cache hits reduce input cost materially.
Highlights
Token Plan Hy tiers use hy3-preview inside tools—self-built backends should use pay-as-you-go API, not plan keys.
Best for
Agent, coding, and long-context production traffic
Usage
Hy-MT2-Pro for translation—¥0.5 input, ¥2 output per 1M tokens.
Models
Suited to localization, cross-border documents, and i18n product pipelines.
Best for
Translation and multilingual content teams
HY-Vision-2.0-Instruct
VisionInput
¥7.5
Output
¥17.5
Usage
HY-Vision-2.0-Instruct at ¥7.5/¥17.5 per 1M—for vision-language understanding and VQA.
Models
Route by task complexity among HY-Vision-1.5-Thinking and HY-Vision-Video.
Best for
Vision-language understanding, VQA, and multimodal agents
HY-Image-V3.0
Text-to-imageInput
¥0.2
Output
image
Usage
HY-Image-V3.0 at ¥0.2/image; HY-Image-Lite at ¥0.099/image for different quality/cost needs.
Models
Image generation bills per image—not directly comparable to language token unit prices.
Best for
Text-to-image, marketing assets, and creative design
Notes
- Hy3 preview tiers: (0,16k] ¥1.2/¥4, [16k,32k) ¥1.6/¥6.4, 32k+ ¥2/¥8 per 1M—cache hits ¥0.4/¥0.6/¥0.8 by band.
- Batch discounts appear for some hosted models on the Guangzhou page—confirm Hunyuan-owned batch rules in the Batch section.
- Singapore region prices differ from Guangzhou—verify TokenHub console rates for your deployment region before launch.
Supported coding tools
OpenAI-compatible APITokenHubBatch inferenceVisionImage generationVideo generation
Pricing and model data sourced from official vendor websites
FAQ
General7 条