ABOUT THIS MODEL

Byte-level text classifier that runs in under 6ms on CPU with a 200KB ONNX export. Classifies 62 programming languages and 90 MIME types. Published on ArXiv (2608.14658: "pico-type: A 1.5M-Parameter Byte-Level Multi-Head Content Classifier"). Designed for edge deployment, air-gapped systems, and high-throughput classification pipelines.

KEY HIGHLIGHTS
  • ✓Published research on ArXiv (2608.14658)
  • ✓Smallest production classifier: 200KB total
  • ✓Sub-6ms inference on any CPU
  • ✓No tokeniser needed — operates at byte level
  • ✓7 classification heads in a single model
SPECIFICATIONS
Parameters1.5M
Inference<6ms on CPU
Binary Size200KB ONNX
Classification62 languages · 90 MIME types
PaperArXiv 2608.14658
LicenseApache 2.0
Downloads44+
QUICK START

Get started in minutes

Install
pip install onnxruntime
ACADEMIC & RESEARCH CITATION

Cite this Model & Architecture

If you utilize pico-type in your academic research, benchmarks, or enterprise deployments, please cite our open-weight publication:

@misc{eulogik2026picotype,
  title        = {pico-type v02: 200KB Edge AI Text Classifier},
  author       = {Kishore, Gautam and Eulogik Systems Engineering},
  year         = {2026},
  publisher    = {Eulogik},
  howpublished = {\url{https://eulogik.com/models/pico-type}},
  note         = {ArXiv: 2608.14658}
}

Need help deploying pico-type?

We offer consulting, custom fine-tuning, and on-premise deployment for all our models.