Looki Unveils Loom Multimodal Model Series
16 hour ago / Read about 0 minute
Author:小编   

Looki has launched Loom, its self-developed multimodal model series designed for first-person life understanding and long-term memory. The model connects the entire pipeline—from continuous perception and experience organization to cross-temporal memory retrieval—weaving fragmented observations into an understandable, traceable personal context. Its related modules deliver outstanding performance: Loom Sense achieves an event-judgment F1 score of 88.6, with a single-window processing time of 0.51 seconds and an inference cost reduction of approximately 94%; Loom Moment outperforms open-source models in segmentation, with more precise evidence attribution; and Loom Memory achieves a comprehensive score of 79 while processing only 14.5% to 30.5% of the information volume required by comparison baselines.