the fusion embedding family
models
one open-weight base holds every modality in a single embedding space. sense packs attach to it without disturbing what already works, and engram turns the whole space into a searchable memory.
fusion embedding 2
flagship · multimodal embeddingone ~2B open-weight model that puts text, image, video, and audio in a single embedding space.
read the case study →fusion embedding 1
foundation · first modelthe first fusion embedding: a frozen VLM base extended with an audio pathway, proving one shared space could hold a new modality.
read the case study →tremor
sense pack · motion / imuaccelerometer and imu motion projected into the fusion embedding space, searchable in plain language.
read the case study →ember
sense pack · thermala thermal sense pack that adds infrared imagery to the shared space at no cost to the other senses.
read the case study →k3 vision
sense pack · second vision heada second, semantics-forward vision tower projected into the fusion embedding space.
read the case study →fusion perception
research · geometrya frozen dino backbone projected into the space, making geometry and place language-searchable.
read the case study →engram
product · memory layeran open cross-modal memory for physical ai, built on the family and searchable in plain language.
read the case study →