Autoregressive MoE Multimodal Model
🎨 Tencent HunyuanImage-3.0-Instruct
Interactive Web Client & Demo for tencent/HunyuanImage-3.0-Instruct (80B MoE Parameters).
🚀 Generate & Edit Images with Instruct
HunyuanImage-3.0-Instruct supports both Text-to-Image generation and Instruction-based Image-to-Image editing with reasoning (Chain-of-Thought).
Your token stays strictly client-side in your browser. Get one at HF Token Settings.
Click to choose or drag & drop an image here
Supports PNG, JPG, WEBP✨ Key Capabilities
🧠 Unified Autoregressive MoE
80B total parameters with 13B activated per token across 64 experts for rich text-visual coherence.
💭 Reasoning & Recaptioning
Leverages native multimodal reasoning to interpret sparse user intent and refine prompt details automatically.
🎨 High-Fidelity Editing
Multi-image fusion, fine-grained object modification, style transfer, and precise instruction following.