On July 27, 2025, at the Tencent Forum of the 2025 World Artificial Intelligence Conference, Tencent unveiled the Hunyuan 3D World Model 1.0 and announced its full open-source availability. This groundbreaking initiative marks the birth of the industry's first open-source world generation model that supports immersive roaming, interaction, and simulation, opening up new possibilities for game development, virtual reality (VR), and digital content creation.

The Hunyuan 3D World Model 1.0 integrates panoramic visual generation with layered 3D reconstruction technology, capable of accepting text and images as inputs to rapidly create high-quality, stylistically diverse, explorable 3D scenes. This technological leap significantly simplifies the 3D scene construction process, reducing tasks that once took professional modeling teams weeks to a matter of minutes with simple text commands or image uploads. For game developers, the model can quickly generate complete 3D scenes with buildings, terrains, vegetation, and more, outputting Mesh files directly usable for game prototyping or level design. It also supports foreground object adjustments and sky background changes, catering to personalized creative needs. Even users without modeling experience can effortlessly create 360° immersive visual spaces through the Hunyuan 3D creation engine and seamlessly import them into virtual headset devices like Vision Pro for an immersive experience.

At the core of the Hunyuan 3D World Model 1.0 lies its innovative "Semantic Hierarchical 3D Scene Representation and Generation Algorithm." This algorithm deconstructs complex 3D worlds into different semantic layers, intelligently separating foregrounds from backgrounds and ground from sky. It not only generates realistic overall scenes but also outputs standardized 3D Mesh assets compatible with mainstream tools like Unity, Unreal Engine, and Blender, facilitating independent editing or physical simulation of scene elements. This seamless integration of AIGC technology with traditional CG workflows sets it apart from global leading open-source models, surpassing them in key metrics such as world generation from text and aesthetics.

Tencent Hunyuan has also disclosed a series of open-source plans, including end-side hybrid inference language models, multimodal understanding models, and game visual models. The flagship model, TurboS, has been rapidly iterating with a new version each month since its release at the beginning of the year, significantly enhancing code generation, scientific capabilities, and complex instruction following, securing its place among the top global large models. Based on the TurboS foundation, Hunyuan T1 has further improved inference capabilities, demonstrating leading comprehensive abilities in China. Additionally, end-to-end voice models like Hunyuan Voice and multimodal understanding models like Hunyuan Vision, built on the Hunyuan large language model base, inherit excellent understanding, cognitive, and reasoning abilities, achieving remarkable results in multimodal understanding.

Tencent Hunyuan is also fully committed to multimodal generation, with its 3D generation capabilities leading globally. The latest Hunyuan 3D2.5 model has significantly enhanced textures, ranking first globally in evaluations by the Shanghai Artificial Intelligence Lab. Besides the Hunyuan 3D World Model 1.0, Tencent will open-source a series of smaller models at the end of the month, including 0.5B, 1.8B, 4B, and 7B hybrid inference models, which are lighter and easier to deploy. The multimodal understanding model Hunyuan-large-vision and the interactive game video generation framework HunyuanGameCraft, optimized for game scenarios, will also be open-sourced soon.

Tencent Hunyuan has opened up its complete multimodal generation capabilities and tool suite plugins, successively open-sourcing industry-leading text-to-image, video generation, and 3D generation capabilities. It provides an open-source base close to commercial model performance, facilitating community customization based on business and usage scenarios. Currently, the community download volume of the Hunyuan 3D series models has exceeded 2.3 million times, making it one of the world's most popular 3D open-source models.

Project Homepage: https://3d-models.hunyuan.tencent.com/world/

Experience Address: https://3d.hunyuan.tencent.com/sceneTo3D

Hugging Face Model Address: https://huggingface.co/tencent/HunyuanWorld-1

Github Project Address: https://github.com/Tencent-Hunyuan/HunyuanWorld-1.0