Xlue
Category: AI in Healthcare
Carnegie Mellon spinout building a Health World Model, a human health foundation model for general clinical intelligence that predicts how patient health evolves and how treatments change it. Xlue was founded in 2025. The company is led by Chenyan Xiong. Based in Pittsburgh, PA, USA. Team size: 1-10. Total funding raised: $36.5M. Latest round: Seed. Key investors include Conviction.
- Founded
- 2025
- Headquarters
- Pittsburgh, PA, USA
- Team size
- 1-10
- Total funding
- $36.5M
Value proposition
Turns clinical questions into computational problems. The model learns from longitudinal patient trajectories, treatment responses and trial data so clinicians and drug developers can predict risk and simulate outcomes before acting.
Products and solutions
Health World Model (human health foundation model), CATCH-FM (cancer pre-screening and risk stratification from EHR/ICD data, initially pancreatic, liver and lung cancer), offline clinical-trial simulation, indication discovery. Planned tasks: drug discovery, patient identification, care planning, precision medicine, diagnostic workup, indication discovery, trial optimization.
Unique value
Pretrains on health states and treatment actions instead of tokens, treating disease progression, treatment response and outcomes as one process of health changing over time. Company-reported blind third-party evaluations claim it beat frontier token-based models at under 0.1% of the compute cost.
Target customer
Hospitals and health systems (preventive care and risk stratification); pharmaceutical and drug-development companies (trial design, indication discovery)
Industries served
Healthcare providers, oncology and preventive care, pharma and biotech drug development
Technology advantage
Health-state abstraction for pretraining on patient trajectories. Training data is large longitudinal health records (Taiwan national claims database, US records). Academic pedigree from CMU, Stanford, Georgetown, Microsoft and NIH. Scaling-law research published on arXiv (2506.00209).
How they differentiate
Models health states and treatment actions rather than tokens, with a recursive loop spanning molecular, clinical-trial and population-health signals, versus EHR-vendor or oncology-data-centric models.
Main competitors
Epic (Comet / Curiosity on Cosmos), Tempus (oncology foundation models), Knit Health, Delphi-2M (academic)
Key partnerships
Kaohsiung Medical University (Taiwan, retrospective 80,000-patient study), in talks with UPMC for de-identified record access, in talks with hospitals and pharma companies, prospective disease-risk studies across multiple health systems
Major milestones
2025: spun out of CMU and incorporated, Sept 2025: Kaohsiung Medical University press release on retrospective cancer studies, Feb 2026: CATCH-FM coverage, $1.5M friends-and-family round, Nov 18, 2026: $35M seed led by Conviction and technical report on the Health World Model
Growth metrics
Six employees as of Feb 2026, before the $35M seed. Retrospective study of 80,000 patients across 7 cancer types flagged 12.6% of cases at least 180 days before diagnosis and 19.5% at least 90 days before.
Market positioning
Early-stage clinical foundation model company, positioned upstream of care. It focuses on risk stratification (not diagnosis) and on pharma trial simulation.
Geographic focus
United States and Taiwan
About Chenyan Xiong
Ex-Microsoft Research Principal Researcher (large language models, text retrieval); Associate Professor, Language Technologies Institute, Carnegie Mellon University School of Computer Science. Co-founder and co-lead of the CMU Foundation and Language Model (FLAME) Center.
Latest news about Xlue
More AI in Healthcare companies
Official website: https://xlue.com