Labops
ORINN
Medical-grade alternative to generic LLMs

High Quality Data

Orinn is trained on high-quality medical data sourced from textbooks, research papers, and partner medical institutions. We don’t rely solely on web scraping.

Specialized Algorithm

No single algorithm is sufficient for every domain, so we designed a highly personalized one specifically for healthcare.

RL with Professionals

We trained and applied RL with a team of doctors, so the model learns to align closely with real medical expertise.

Affordable Intelligence

Our new model, Orinn 1.5, is far more affordable than others—hosting 8B model while delivering the same level of efficiency.

HealthBench

We evaluated Orinn 1.5 on OpenAI’s HealthBench, where it achieved a score of 0.521—slightly ahead in a direct comparison with Google’s Gemini 2.5 Pro, which scored 0.520.

By plotting reliability curves for multiple models, we observed that Orinn 1.5 delivers a notable improvement in worst-case performance compared to earlier generations, while still leaving space for further progress.

Emergency
Expertised
Uncertainty
Response
Health
Global
Context

QA Benchmark

We also evaluated Orinn 1.5 on leading medical QA benchmarks, including MedMCQA, MedQA, PubMedQA, HealthBench, and medical-related questions from MMLU-Pro, where it delivered impressive results, outperforming Google’s MedGemma-27B-IT.

Pub Med
Med M C
Med Q A
M M L U_ P
Health Bench

Every thought has a path.
Orinn follows it.

LLMs generate answers. Orinn maps the reasoning behind them — with context, clarity, and care.

Prompt
|

Orinn 1.5 is here

The next era of healthcare starts here.

Join us in scaling Orinn 1.5 — the AI health companion designed to serve billions, not just millions. We’re building a future where clinical-grade intelligence lives in every home.

Invest in trust. Invest in Orinn.