Tempus AI (NASDAQ: TEM) announced initial results from its multimodal oncology foundation model at the 2026 American Society of Clinical Oncology (ASCO) Annual Meeting.
The transformer-based model utilizes a dataset of 2.5 million patient records. It incorporates 250 million pages of clinical notes. The system also processes hundreds of thousands of medical images and genomic sequences.
The model employs zero-shot learning to demonstrate predictive capabilities without additional fine-tuning. In a non-small cell lung cancer study, the AI achieved high accuracy in predicting overall survival. It successfully stratified patients into high- and low-risk groups.
Tempus stated the model outperformed standard predictive approaches in mirroring major clinical trial outcomes.