On Oct. 5, 2025, Tencent’s newly open-sourced Hunyuan Picture 3.0 has vaulted to the highest spot on the LMArena text-to-image leaderboard, edging out high-profile proprietary techniques together with Google’s “nano-banana” (Gemini 2.5 Flash Picture Preview). The rating is predicated on LMArena’s human “blind-vote” battles by which customers choose most well-liked outputs with out realizing which mannequin generated them.
Hunyuan Picture 3.0 is positioned as a local multimodal, autoregressive mannequin aimed toward each understanding and era; Tencent launched the mannequin with weights and code below an open-source license on Hugging Face. The mannequin card describes it as a big MoE system and highlights text-to-image high quality that “matches or surpasses” main closed fashions in inner assessments.
LMArena’s newest Textual content-to-Picture Enviornment web page lists “hunyuan-image-3.0” at No.1 as of Oct. 3, 2025 (leaderboard timestamp), whereas the Enviornment’s official social account and Tencent Hunyuan’s account additionally famous the leap. As a result of the mannequin is new on the platform, its vote rely is presently far decrease than long-standing entries; the rating is subsequently marked “Preliminary.”
The Hunyuan group has been increasing its open stack in latest weeks, publishing associated 3D property and examples alongside the picture mannequin. Builders can obtain Hunyuan Picture 3.0 straight from its Hugging Face repository, and a number of other group demo areas are already dwell.
Elevate your perspective with NextTech Information, the place innovation meets perception.
Uncover the newest breakthroughs, get unique updates, and join with a world community of future-focused thinkers.
Unlock tomorrow’s traits immediately: learn extra, subscribe to our publication, and develop into a part of the NextTech group at NextTech-news.com
