Tencent has launched HunyuanImage 2.1, its newest text-to-image model. The interesting part is that the company has also open-sourced the new model and is said to improve image quality, style variety, and text rendering for creators. It brings native 2K resolution generation, faster performance, and enhanced control over complex prompts.
HunyuanImage 2.1 comes with some important upgrades. For example, it now supports ultra-long prompts up to 1000 tokens, allowing creators to describe highly detailed scenes or multiple subjects in one image. Its precise Chinese and English text rendering integrates text seamlessly into visuals, making it practical for uses like posters, product covers, and illustrations. Besides, the model also offers diverse artistic styles from photorealistic portraits to vinyl figures while maintaining high aesthetic quality.
Under the hood, the model uses a 17-billion-parameter dual-stream diffusion transformer. It combines a multimodal large language model to better align text and image generation with a multi-language, character-aware encoder for more accurate text handling. Tencent has also released the weights for an accelerated version using meanflow, which cuts inference steps from 100 to just eight, and PromptEnhancer, a rewriting tool that refines prompts for more expressive outputs. These additions are designed to help users generate visuals, like slogan-heavy posters or multi-panel comics, faster than ever.
However, this is not the only update from the house of Tencent. Recently, the company launched its Hunyuan-MT-7B, a translation model that dominated the WMT2025 machine translation competition. It also launched its Hunyuan-MT-Chimera-7B, which is an integrated translation system that blends outputs from multiple models for more accurate results, especially in specialised fields.
Furthermore, the company has hinted that it’s working on something more by teasing a native multimodal image generation model as the next milestone in its roadmap. The heat in the AI-backed image editing or text-to-image space is heating up with Google’s recent updates. But by making HunyuanImage 2.1 open-source, the company is positioning itself as a serious player in this competitive market.
