Developer API
Build with the most advanced clinical artificial intelligence
Integrate Clinical‑Grade, Evidence‑Based Intelligence
With the Glass API, you can embed clinical-grade, evidence-based AI into your products like electronic health records, telehealth platforms, and clinical workflow apps - without the burden of building, evaluating, and updating your own clinical AI system. Unlock new clinical AI use cases in your workflows and products including triage, patient record summarization, diagnostic support, treatment planning, automated documentation, and ambient scribing.
Agentic AI Architecture
Glass is a large language model-based AI agent powered by sophisticated medical knowledge retrieval to deliver trustworthy and accurate evidence-based responses. Purpose-built for healthcare, Glass processes clinical queries, analyzes patient data, and returns guideline-aligned recommendations with cited sources.
Core capabilities include:
Evidence-Based Q&A
Ask any clinical question and receive authoritative answers backed by the latest guidelines, trials, and medical resources in seconds.
Patient Data Summarization
Generate concise, structured summaries of EHR data (notes, labs, imaging, medications, history).
Differential Diagnosis Generator
Expand diagnostic thinking with AI-ranked differentials and diagnostic next steps.
Treatment Plan Generator
Craft guideline-aligned management plans tailored to patient context.
Documentation Generation
Create comprehensive clinical documentation—progress notes, discharge summaries, referrals, prior authorization letters, insurance appeals, and more.
Ambient Scribing
Generate diarized transcripts and structured clinical documentation from encounter audio through the Scribing API endpoint.
Real-Time Awareness of the Latest Medical Literature
Glass indexes the medical literature from a variety of sources including more than 38 million peer-reviewed articles and comprehensive FDA-approved drug information covering 154,000+ compounds. Our AI agent searches these knowledge bases with sophisticated analysis - mirroring the discernment of a medical scholar while applying clinical guidelines with expert-level judgment.
Clinical Artificial Intelligence Benchmark Suite
Comprehensive internal model scorecard evaluating clinical AI accuracy and competencies across nine benchmarks spanning medical knowledge recall, multi-step reasoning, diagnostic reasoning, hallucination detection, evidence appraisal, and structured clinical documentation.
| Benchmark | Glass 5.5 Deep Reasoning |
Glass 5.5 Standard |
Opus 4.6 | GPT-5.4 | Sonnet 4.6 | GPT-5.2 | Opus 4.5 | Sonnet 4.5 |
|---|---|---|---|---|---|---|---|---|
| Overall | 88.6% | 85.1% | 81.8% | 81.4% | 80% | 79.2% | 78% | 76.9% |
| NEJM CPCs | 84% | 82% | 54% | 72% | 47% | 64% | 62% | 64% |
| AMEGA | 96% | 88% | 84% | 79% | 92% | 74% | 77% | 70% |
| JAMA CCS | 93% | 86% | 81% | 85% | 84% | 82% | 81% | 80% |
| MedXpertQA | 55% | 47% | 48% | 46% | 44% | 47% | 43% | 39% |
| PubMedQA | 79% | 79% | 77% | 75% | 71% | 66% | 77% | 72% |
| MedQA | 97% | 94% | 97% | 93% | 92% | 92% | 82% | 90% |
| MMLU-Med | 98% | 95% | 96% | 94% | 97% | 95% | 86% | 91% |
| MedHallu | 96% | 96% | 98% | 92% | 95% | 93% | 94% | 88% |
| ACI-Bench | 100% | 99% | 100% | 97% | 98% | 100% | 100% | 98% |
Internal benchmarking carried out against the Glass Developer API and frontier model APIs, evaluating model responses against queries from established external clinical AI evaluation datasets. All models evaluated using identical prompts and graded against correct answers as defined by the external benchmarks.
Leading Performance on Clinical Benchmarks
Our clinical intelligence is rigorously evaluated against clinical benchmarks and validated by academic clinicians. Clinicians worldwide rely on Glass for best-in-class AI clinical decision support, including clinicians at leading healthcare systems like Harvard Medical School, Stanford Medicine, and Johns Hopkins Medicine.
Technical Specifications
The Glass API provides a RESTful interface with secure authentication and structured endpoints for clinical queries. It returns markdown-formatted responses with in-text citations and provides all information and metadata for the citations used for reference display. The system delivers real-time progress updates from the agent during processing, which can be displayed directly to users.
Built for Scale and Reliability
Reliability at Scale
Typical error rates are below 0.01% via intelligent fallback mechanisms.
Elastic Deployment
Autoscaling, load-balanced servers across regions.
Build with the World's Most Advanced Evidence-Based Clinical AI
Integrate leading clinical intelligence capabilities - including differential diagnosis, treatment planning, clinical documentation, and ambient scribing - into your product through a single API.