← Founder Notes
Archive

Google's open models just went edge-native. gemma 4, out september 16, runs audio and vision fully…

Yethikrishna ROriginal on Threads

google's open models just went edge-native. gemma 4, out september 16, runs audio and vision fully offline on phones, raspberry pi and jetson, with e2b and e4b sizes.

frontier ai now has a pocket version.

Context

Google's Gemma 4 launch post, dated April 2, 2026, lists four sizes, Effective 2B (E2B), Effective 4B (E4B), 26B Mixture of Experts and 31B Dense. It says all models natively process video and images and that E2B and E4B add native audio input for speech recognition. It says the multimodal models run completely offline with near-zero latency on phones, Raspberry Pi and NVIDIA Jetson Orin Nano.

How it compares

The note says out September 16. The launch post is dated April 2, 2026 and no September 16 release page was found. Audio is native on E2B and E4B only. The offline and latency statements are Google's and not independently benchmarked. The Android route named is the AICore Developer Preview, which is a preview. The license name was not extracted.

Watch next

  • Google's Gemma release notes for any later update, and independent on-device benchmarks.

Sources

  1. Google: Gemma 4 (April 2, 2026)blog.google

Provenance

The note above is reproduced unedited from the original post, first published on Threads on 21 September 2026 at 19:18 IST. Sources are the papers and datasets the note draws on.

View the original post
Embed this note
<iframe src="https://founder.myndlabs.tech/notes/embed/google-s-open-models-just-went-edge-native-DdjTdRYjRWb" width="480" height="420" style="border:0;max-width:100%" loading="lazy" title="Google's open models just went edge-native. gemma 4, out september 16, runs audio and vision fully…"></iframe>

More notes