Model releases  ·  11 August 2026

Gemma 4 QAT Mobile: builds made for phones, not shrunk for them

Google published separate quantisation-aware-trained variants of its two smallest Gemma 4 models, aimed at mobile deployment — a different thing from taking the standard weights and compressing them afterwards.

At a glance

Released byGoogle DeepMind
Published20 July 2026
VariantsE2B and E4B, each in -ct and -transformers form
ArchitectureGemma4ForConditionalGeneration (gemma4)
Context length128K on the small models, 256K on medium
ModalitiesText and image; audio supported on E2B, E4B and 12B
LanguagesMore than 140
LicenceApache 2.0
Weight size2.67 GB for E2B QAT mobile
Hosted atgoogle on Hugging Face

What QAT actually changes

Most small models reach a phone through post-training quantisation: the lab publishes full-precision weights and someone compresses them afterwards, accepting whatever quality is lost. Quantisation-aware training puts the compression inside the training loop, so the model learns weights that survive it.

The practical difference is that a QAT four-bit build is generally closer to its own full-precision parent than a post-hoc four-bit build of the same model. Google publishing these as separate repositories, rather than as a note in the main model card, is a signal about how it expects Gemma 4 to be deployed.

Architecture

Gemma 4 interleaves local sliding-window attention with full global attention, with the final layer always global. Google's framing is that this gives the speed and memory profile of a lightweight model while keeping long-context awareness. To hold memory down at long context, the global layers use unified keys and values and apply Proportional RoPE.

The family is multimodal — text and image input across the line, with audio supported on E2B, E4B and 12B — and covers more than 140 languages with native support for a system role.

What's published

Four repositories: E2B and E4B, each in a -ct and a -transformers flavour. The E2B QAT mobile weights are 2.67 GB. All are Apache 2.0, which puts them among the more permissively licensed models available in this size class.

Runtime support

RuntimeSupportedNotes
llama.cppYesThe gemma4 architecture has upstream support and community GGUF builds are widely available.
transformersYesThe -transformers variants are published for exactly this path.
MLXPartialCommunity conversions; nothing published by Google.
Audio inputPartialSupported on E2B, E4B and 12B, but runtime support for audio lags behind text and image.

What we don't know yet

Sources

Frequently asked questions

What is quantisation-aware training?

Training that includes the compression step in the loop, so the model learns weights that survive being reduced to four bits. It generally holds quality better than compressing full-precision weights after training.

What licence is Gemma 4 under?

Apache 2.0, including the QAT mobile builds.

Does Gemma 4 support images and audio?

Text and image across the family, with audio supported on the E2B, E4B and 12B sizes. Runtime support for audio varies by engine.

How large is the Gemma 4 E2B QAT mobile build?

2.67 GB.

What context length does Gemma 4 support?

128K tokens on the small models and 256K on the medium ones.

OnDevice LLM is a private AI assistant that runs entirely on your iPhone — no account, no cloud, and nothing you type leaves the device.

Published 11 August 2026