Tech companies advance on-device AI memory and multimodal hardware

Several technology companies are advancing on-device AI and specialized hardware. OPPO, in collaboration with the OpenKG community and various universities, has launched MobileMem, the industry’s first evaluation benchmark for on-device AI memory. The benchmark focuses on long-term event modeling, cross-modal reasoning, and personalized adaptation to help developers standardize AI memory capabilities.

In the smartphone sector, vivo is expanding its Atomic Workbench feature from foldable devices to standard slab phones via OriginOS 7. This interface allows up to five applications to run in a continuous workflow using proprietary scheduling and memory management algorithms. Similarly, MemVerge founder Charles Fan emphasizes that AI memory ownership will become a critical competitive moat as models become increasingly homogenized.

In consumer electronics, Taiwan-based OVO has unveiled several new products, including the TT9 4K mobile screen which integrates Google Gemini Live for multimodal visual interaction. Other releases include the SS2 laser short-throw projector with real-time vocal removal for karaoke, and the Warpple series of portable outdoor projectors and speakers.