MAI-Thinking-1

This 15-billion-parameter video model (developed by Alibaba) took first place in the Artificial Analysis rankings for video generation, ahead of Seedance 2.0. It generates text-to-video and image-to-video clips in 1080p with lip-sync in seven languages, all within a single unified pipeline

A 3D model that generates interactive 4D objects (cars with spinning wheels, automatic scripts) from textual instructions. Tokenization of 3D shapes, generation of shapes/scenes from text, and real-time character animation

An AI architecture with integrated long-term memory, selective updating based on a surprise signal, and context extended beyond 2 million tokens. MIRAS unifies transformers and linear RNNs thanks to optimized associative memory

Run your GPU workloads in a confidential environment protected by Intel TDX (with cryptographic proof of integrity). Confidently deploy certified TDX images, encrypted storage, and hardware isolation