Show HN: AviGPT-250M – Edge SLM with NVMe memory bus (488MB, runs offline)

Category: ai-ml

Tags: edge-ai, small-language-model, retrieval-augmented-generation, sqlite, semi-parametric

Score: 5.0/10 (Innovation: 6, Technical: 5, Documentation: 5, Utility: 4)

AviGPT-250M is a 250M-parameter small language model for edge devices that augments the parametric model with a semi-parametric 'NVMe memory bus' using SQLite FTS5 for factual retrieval and an AST-based calculator for deterministic math. The project claims zero hallucination on indexed knowledge and high efficiency, though benchmark methodology (self-defined composite efficiency) and competitive claims appear loosely validated.

Target audience: edge AI developers, ML engineers, embedded systems developers

Repository: https://github.com/Avinashricky211/AviGPT-250M · Python · 4 stars

View on Hacker News