← The Wire

Models2026-06-08 · source-backed

Google shipped Gemma 4 12B (Apache 2.0) plus QAT variants for on-device.

Confidence
source-backed
Sources
2
Redaction
redacted

Story

Google released Gemma 4 12B June 3 under clean Apache 2.0, native multimodality, up to 256K context on larger variants, with the 31B reportedly at 85.2% MMLU Pro. Two days later came QAT versions optimized for mobile and laptop hardware. The 12B size targets the single-GPU sweet spot, and QAT preserves accuracy at lower bit-widths, which makes local agentic workloads actually practical instead of aspirational.

Related stories

Each link below shares sources, entities, or timing with this story.

  1. Shared entities / Shared topic / Earlier coverage

    Google shipped a 12B multimodal model that runs on my laptop and hears audio with no encoder

    Both cover Apache, Gemma, Google, GPU; overlapping topics (gemma, google); earlier Apache coverage from 2026-06-04.

  2. Gemma 4 Drops Four Open Models, and the 31B Dense Beats a 397B MoE

    Both cover Apache, Gemma, Google; overlapping topics (apache, context, gemma, google); earlier Apache coverage from 2026-04-05.

  3. Google Ships Gemma 4 Under Apache 2.0. Four Models, Arena #3, and the License That Changes Everything.

    Both cover Apache, Gemma, Google; overlapping topics (apache, gemma, google, variant); earlier Apache coverage from 2026-04-02.

  4. Gemma 4: 1,200% Improvement in Agentic Tool Use.

    Both cover Apache, Gemma, Google; overlapping topics (agentic, apache, gemma); earlier Apache coverage from 2026-04-21.

  5. Gemma 4 Ships Under Apache 2.0 and the Numbers Are Absurd

    Both cover Apache, Gemma, GPU; overlapping topics (apache, gemma, variant); earlier Apache coverage from 2026-04-13.

  6. Gemma 4 31B Just Made Half Your API Budget Obsolete

    Both cover Gemma, Google, MMLU Pro; overlapping topics (actually, gemma, google); earlier Gemma coverage from 2026-04-06.

  7. Shared entities / Shared topic / What happened next

    Google open-sourced DiffusionGemma, a text-diffusion model that generates 4x faster, and you can run it locally today

    Both cover Apache, Gemma, Google; overlapping topics (actually, google); picks up the Apache thread on 2026-06-11.

  8. Shared entities / Shared topic / Earlier coverage

    Google Ships Multi-Token Prediction for Gemma 4. Local Inference Just Got 3x Faster.

    Both cover Apache, Gemma, Google; overlapping topics (gemma, google); earlier Apache coverage from 2026-05-06.

Source trail

Entities

Provenance

AI generated
no
Story unit
2026-06-08-google-shipped-gemma-4-12b-apache-2-0-plus-qat-variants-for-on-device
Labels
source-backed, canonical briefing excerpt