inference/research
verifiedhigh · 0.90updated 2026-10-03

Gemma 3n model overview

spec-sheet

Fields

id
sources/model-fr-google-gemma-3n-doc
URL
https://ai.google.dev/gemma/docs/gemma-3n
Publisher
Google
Kind
spec-sheet
Published
2025-06-03
Accessed
2026-10-03
notes
Documents PLE (Per-Layer Embedding) caching to fast storage, MatFormer nested-sub-model activation, and conditional parameter loading of the vision/audio towers. States the E2B loads over 5B parameters in standard execution but can run with an effective 1.91B load.

Notes

Documents PLE (Per-Layer Embedding) caching to fast storage, MatFormer nested-sub-model activation, and conditional parameter loading of the vision/audio towers. States the E2B loads over 5B parameters in standard execution but can run with an effective 1.91B load.

Cross-references

This record references no other records.