HunyuanImage-3.0

HunyuanImage-3.0 is a native multimodal model designed for text-to-image generation, operating within an autoregressive framework that unifies multimodal understanding and generation. It achieves performance comparable to leading closed-source models, emphasizing semantic accuracy and photorealistic visual quality.

Key features:

• Unified Autoregressive Architecture: Integrates text and image modalities directly, enabling contextually rich generation without reliance on DiT-based architectures.
• Large-Scale MoE Model: Features 80 billion total parameters (13 billion activated per token) with a Mixture of Experts (MoE) configuration of 64 experts.
• Advanced Reasoning: Leverages world knowledge for interpreting user intent, enhancing sparse prompts with contextual details.
• Supported Tasks: Text-to-image generation, including complex prompts and multi-turn interactions. 
• System Requirements: 4×80GB recommended GPU memory, disk space: ~170GB for model weights, Python 3.12+, PyTorch 2.7.1.

Announce Date: 25.09.2025
Parameters: 80B
Experts: 64
Activated at inference: 13B
Attention Type: Full Attention
Developer: Tencent
vLLM Version: 0.16.0
License: tencent-hunyuan-community

Use our pre-built public endpoints for free to test inference and explore HunyuanImage-3.0 capabilities. You can obtain an API access token on the token management page after registration and verification.

Rent your own physically dedicated instance with hourly or long-term monthly billing.

We recommend deploying private instances in the following scenarios:

• maximize endpoint performance,
• enable full context for long sequences,
• ensure top-tier security for data processing in an isolated, dedicated environment,
• use custom weights, such as fine-tuned models or LoRA adapters.

Contact our dedicated neural networks support team at nn@immers.cloud or send your request to the sales department at sale@immers.cloud.