AI Series #49: FORCE Conference Next Week; Weixin AI; Fast Model iterations

Research evidence excerpt

AI Series #49: FORCE Conference Next Week; Weixin AI; Fast Model iterations

ade in vision-language

capabilities while retaining full agentic strength in coding, tool use, and productivity workflows.

.

Source: Company, Jefferies

Exhibit 89 - Tencent model list

Release Date Model Core Functions

May-24 Hunyuan-DiT The industry's first Chinese-native Diffusion Transformer (DiT) for high-quality text-to-image generation.

Sep-24 Hunyuan-Turbo An upgraded LLM based on the Mixture-of-Experts (MoE) architecture, doubling training efficiency and reducing

inference costs.

Nov-24 Hunyuan3D 1.0 First open-source 3D generation model supporting text-to-3D and image-to-3D workflows.

Dec-24 HunyuanVideo A large-scale video generation framework capable of producing high-definition videos with complex motion.

Jan-25 Hunyuan3D 2.0 Significant upgrades in geometry accuracy and texture representation for 3D asset creation.

Feb-25 Hunyuan Turbo S A "fast-thinking" model using Hybrid-Mamba-Transformer architecture, optimized for near-instant responses

and competitive reasoning.

Mar-25 HunyuanVideo-I2V An specialized image-to-video model built on the foundation of HunyuanVideo.

Jun-25

Hunyuan3D 2.1 Production-ready 3D asset generation with a new PBR (Physically Based Rendering) material model.

Sep-25 Hunyuan Image 3.0 Flagship image generation model (80B parameters) topping global rankings for semantic understanding.

Nov-25 HunyuanVideo 1.5 A highly efficient and powerful update to the foundation video generation model.

Dec-25 Hunyuan 2.0 Next-generation LLM featuring 406B parameters and a 256K context window for complex reasoning tasks.

Jan-26 Hunyuan Image 3.0-Instruct An image-to-image editing model supporting precise single-sentence editing and multi-image fusion.