The Agent RaceJuly 10, 2026via MarkTechPost
Google Research Introduces SensorFM: A Wearable Health Foundation Model Pretrained on One Trillion Minutes of Sensor Data
Why it matters
Google Research demonstrates a new frontier in foundation models: scaling laws apply beyond text and images to unlabeled sensor signals. SensorFM's pretraining on 1T minutes from 5M participants and its performance across 34/35downstream health tasks signals a shift toward multimodal, real-world health AI that could reshape clinical workflows and consumer wearables.
Key signals
- Pretraining dataset: 1 trillion minutes of unlabeled sensor signals
- Participant pool: 5,000,000 consented participants
- Downstream task performance: 34 of 35 tasks beat feature-engineered baselines
- Model backbone: ViT-1D masked-autoencoder architecture
- Co-scaling tested across 4 model sizes and 4 data volumes
- Agentic system searched 30,516 prediction heads
- Clinical grounding: Personal Health Agent evaluated by clinicians
- Collaborators: Google Research, Google DeepMind, university partners
The hook
One trillion minutes of sensor data. Google just released SensorFM—a foundation model that redefines what's possible in wearable health AI.
SensorFM, a wearable health foundation model from Google Research, Google DeepMind, and university collaborators. We walk through its ViT-1D masked-autoencoder backbone, pretrained on more than one trillion minutes of unlabeled sensor signals from 5,000,000 consented participants. We examine the co-scaling results across four model sizes and four data volumes, including the case where capacity outruns data. We show how frozen embeddings plus a PCA-50 linear probe beat feature-engineered baselines on 34 of 35 tasks. We also review the agentic classroom that searched 30,516 prediction heads, and the clinician evaluation grounding a Personal Health Agent.
The post Google Research Introduces SensorFM: A Wearable Health Foundation Model Pretrained on One Trillion Minutes of Sensor Data appeared first on MarkTechPost.