Cura 1T: Specialized LLM for Agentic Healthcare Tasks

Healthcare demands AI models capable of handling complex tasks like patient consultations, clinical reasoning over diverse data (text and images), interactive diagnosis, and integration with electronic health records (EHR). Existing specialized LLMs often struggle to cover this full spectrum of use cases without performance degradation in specific areas. Cura 1T is introduced as a healthcare-specialized LLM specifically designed for agentic healthcare applications. Its development employs a novel human-gated self-evolution loop, where a training agent plans target capabilities, trains the model, evaluates performance against benchmarks, and refines the data mixture based on observed failures. This data-centric approach uses targeted synthetic and curated examples to improve the model. Evaluations across a comprehensive healthcare suite show Cura 1T performing at or near the top among leading baseline models. Crucially, it maintains competitiveness on out-of-domain reasoning and agentic benchmarks, demonstrating its robustness and versatility for critical healthcare workflows.