@MaziyarPanahi: We got 755 tokens per second! That's OpenMed privacy-filter v2 (nemotron, MLX 8-bit) reading a 13,000-token clinical fi…
Summary
OpenMed privacy-filter v2 using nemotron and MLX 8-bit achieves 755 tokens per second on a Mac, redacting 1,152 PII identifiers across 22 categories from a 13,000-token clinical file without data leaving the machine.
View Cached Full Text
Cached at: 07/04/26, 08:53 PM
We got 755 tokens per second! 🔥
That’s OpenMed privacy-filter v2 (nemotron, MLX 8-bit) reading a 13,000-token clinical file and redacting every identifier as it streams by. 1,152 caught across 22 PII categories, on a Mac. Nothing left the machine. https://t.co/SeoCvSIiGA
Similar Articles
@MaziyarPanahi: 6,000,000+ PyPI downloads in under a year. OpenMed 1.5.5 ships today: batch PII, on-device redaction in 9 languages, pl…
OpenMed 1.5.5 ships with batch PII redaction on-device in 9 languages, open-source under Apache 2.0, having surpassed 6 million PyPI downloads in under a year.
maziyarpanahi/openmed
OpenMed is an open-source local-first healthcare AI toolkit that provides entity extraction, PII de-identification, and over 1,000 specialized medical models, all running on-device with no cloud dependency.
650+ Apache-2.0 biomedical NER/de-id models that run on-device in MLX. Same fp32 weights, identical outputs: the clinical NER models run 30-40x faster than PyTorch-CPU on a 3-year-old M3 Max. Repro inside.
A collection of 650+ Apache-2.0 licensed biomedical NER and de-identification models that run on-device via MLX, achieving 30-40x faster inference than PyTorch-CPU on an M3 Max with identical outputs.
OpenMed 1.8: Apache-2.0 clinical de-identification that runs fully local, now on Android, iOS, and in the browser. 400+ open issues if you want in on 1.9
OpenMed 1.8 is an Apache-2.0 toolkit for clinical de-identification that runs entirely locally, with new support for Android, iOS, and browser platforms, and invites community contributions for version 1.9.
@MaziyarPanahi: A year ago, OpenMed didn't exist. Today: 340M model downloads. 1,500+ open medical models, all Apache 2.0. 650+ run on …
A year after its inception, OpenMed has achieved 340 million model downloads, offering over 1,500 open medical models under Apache 2.0, with 650+ capable of running on-device on iPhones.