Gemma 4 model overview · inference-research

Fields

id

sources/model-fr-google-gemma-4-doc

URL

https://ai.google.dev/gemma/docs/core

Publisher

Google

Kind

spec-sheet

Published

2026-07-08

Accessed

2026-10-03

notes

Documents the four Gemma 4 architectures (small PLE-based, dense 31B, MoE 26B A4B, unified 12B), the per-size inference memory table, and that every Gemma 4 size ships an MTP draft model for speculative decoding. States all 26B of the A4B model must be resident for fast routing.

Notes

Documents the four Gemma 4 architectures (small PLE-based, dense 31B, MoE 26B A4B, unified 12B), the per-size inference memory table, and that every Gemma 4 size ships an MTP draft model for speculative decoding. States all 26B of the A4B model must be resident for fast routing.

Cross-references

This record references no other records.