Download
Prevon models
Tick the box to show the download links.
Prevon Pulse MK6
The model that picks which action to run in the Prevon app, from what you type. Built for Prevon's own prompt format: it is not a general chat model.
Built on google/functiongemma-270m-it. Weights, training notes and test results on Hugging Face.
- Accept the terms first
prevon-pulse-mk6.litertlm
LiteRT-LM: Android, iOS, Apple Silicon Mac, Windows, Linux · 457 MB
SHA-256
0bd8ddf568a5044bcc35384308d92a43bb8e5b5202c5528e823844251a75f53d - Accept the terms first
prevon-pulse-mk6-Q8_0.gguf
GGUF Q8_0: llama.cpp, Intel Mac · 300 MB
SHA-256
b144aeea25629dc59bf44c2d9c10a51ed91471dcdabeedbea171904c8d958a63
Always included with a model: NOTICE, PREVON_MODEL_TERMS.md, GEMMA_TERMS_OF_USE.txt, SHA256SUMS. Keep them with the model if you share it.
Who may use them
People, schools, researchers and non-profits can use the models free of charge. A company, or anything that makes money, needs our written permission, and competitors are never covered. The Prevon Model Termsare eleven short lines; they sit on top of Google's Gemma Terms of Use, because the models are fine-tuned versions of Google's Gemma. To ask for permission, use the contact details on that page.
Check what you downloaded
Every file has a SHA-256 hash on this page, on Hugging Face, and inside the Prevon app (the app refuses a file that does not match). To check one on a Mac or Linux machine, run shasum -a 256 <file> and compare; on Windows, Get-FileHash <file>. Each model also has a SHA256SUMS file next to it.
Where the files come from
The files are served from downloads.prevon.ai, our own storage on Cloudflare. You do not need an account, and the host keeps no request logs of its own. The Prevon app does not use this page: it downloads the same files from Hugging Face, as described in the Privacy Policy. These models are made for the Prevon app's own prompt format and can be wrong, so check what they produce before you rely on it.
