Tencent Holdings’ latest artificial intelligence model, Hunyuan Image 3.0, has taken the top spot on a major global leaderboard for text-to-image generation, surpassing Google DeepMind’s Nano Banana model.
According to rankings on LMArena, an AI model evaluation platform founded by researchers at the University of California, Berkeley, Hunyuan Image 3.0 now leads both open-source and closed-source competitors.
The model, released in late September, is described by Tencent as “completely comparable to the industry’s flagship closed-source models.” With 80 billion parameters, it is currently the largest open-source image-generation model available.
Parameters are data points that define how a model interprets and generates information — the higher the number, the greater the potential complexity and quality, though it also demands more computing power.
Hunyuan Image 3.0’s rise follows the rapid success of Google DeepMind’s Gemini 2.5 Flash Image, known as Nano Banana, which gained attention in August for its advanced image-editing precision and ability to create detailed 3D figurines.
Tencent’s achievement marks a major milestone for China’s growing AI sector, as open-source innovation continues to challenge the dominance of Western AI labs.
