Perplexity open-sources its PII classifier for hybrid compute, a ~596M-param Qwen3 model
AravSrinivas · x · 2026-09-01
Perplexity open-sourced the PII classifier that decides when hybrid compute routes a workload to the local model. Available on Hugging Face as perplexity-ai/pplx-pii-masking, the model is built on Qwen3 (596M parameters, mostly BF16), MIT-licensed, multilingual, and performs token classification for privacy/PII masking before data leaves the device.
Related event: Perplexity Brings Hybrid Compute to Mac App and Open-Sources PII Classifier(9 posts)→
More from Models
- Gemini 3.7 Flash speedruns Pokemon Kanto by writing code to simulate physics — DynamicWebPaige · 2026-09-01
- Gemini 3.7 Flash speedruns Pokemon via code execution — DynamicWebPaige · 2026-09-01
- Users Report Claude Coding Failures: "I Don't Know What to Do" — doooyle · 2026-09-01
- 0.8B Finetuned Model Beats GPT-5.6 on Specialized Task — sachdh · 2026-09-01
- Alleged deceptive quantization: AtomicChat accused of faking Q4 levels — po_stulate · 2026-09-01
- New PACT Benchmark Reveals Enterprise AI Compliance Failures Under Pressure — baseten · 2026-09-01