micro1

Today we’re announcing the micro1 Medical Index: a suite of medical benchmarks built and graded by working doctors, biostatisticians, regulatory experts, and psychiatrists. It brings all of our medical research into one place, with a combined score across models and a living leaderboard that grows as we publish new benchmarks. We’re starting with three, each built around a fundamental question: - RealmPathology Benchmark: can a model reason over a real pathology report, staying exact on what it says, within its diagnostic limits, and without escalating beyond what the evidence supports? - PharmaBench: can a model work through the regulatory and clinical documentation behind an FDA drug program with the precision a clinical development team would expect? - WellnessBench: when a young person brings a sensitive health question, can a model stay safe without shutting the conversation down, and hold that line under pushback? Together, these three benchmarks cover a meaningful range of medical work, from diagnostic reasoning to pharmaceutical research and patient safety. While they don't yet capture the full scope of the entire medical industry, they establish a strong starting point that will become more comprehensive as we expand the index. Huge shoutout to every expert who built and graded these tasks. Full results linked in the comments below!

We’re hosting micro1’s global Eval Hackathon. 9 days. online. individual. free to join. You’ll use your expertise to build an eval that tests where today’s frontier AI models still fall short, across software, finance, legal, healthcare, science, engineering, and more. ✅ $10,000 in cash prizes ✅ Top performers will be considered for opportunities with micro1 Think you know what AI still can’t do? Build an eval and prove it.

Today we’re launching flow-transform 1.0, micro1’s PII transformation model, delivering frontier-level performance across detection, identity synthesis and full-corpus transformation. Enterprise operational data is uniquely valuable for training frontier AI models because it captures years of how work actually happens. But that data is deeply intertwined with real-world identities. Traditional redaction can remove the PII, but it often breaks the context and relationships that make the data valuable for training. flow-transform 1.0 solves this by creating a privacy-preserved digital twin of the entire enterprise. Rather than anonymizing records independently, the model replaces each real-world entity with one consistent synthetic identity across the dataset. The identities change while the underlying business relationships and workflows stay intact. On PrivacyBench, flow-transform 1.0 reaches 96.0% F1, outperforming every detection baseline we tested. We’re also introducing the Enterprise De-Identification Bench and Transformation Quality Index (TQI) to evaluate the quality of the transformed dataset, not just whether PII was detected. Read the full paper: https://lnkd.in/gRrjyDeM

We’re hosting micro1’s first global hackathon with HackerEarth from August 28–30. 3 days. online. individual. free to join. You’ll use coding agents to tackle a real-world software engineering problem where incomplete requirements, hidden dependencies, and edge cases will test your actual engineering judgment. ✅ $5,000 in cash prizes ✅ Up to 50 paid opportunities with micro1 Can’t wait to see what you build!

We’re hosting micro1’s first global hackathon with HackerEarth from August 28–30. 3 days. online. individual. free to join. You’ll use coding agents to tackle a real-world software engineering problem where incomplete requirements, hidden dependencies, and edge cases will test your actual engineering judgment. ✅ $10,000 in cash prizes ✅ Up to 50 paid opportunities with micro1 Can’t wait to see what you build!

Leverage your expertise to train next-generation AI models