USASI
Model release

Gemma 4 12B Unified

Release in the Gemma family · version 4 (12B Unified)

Maintained by Google DeepMind14

Gemma 4 12B Unified is an 11.95-billion-parameter Gemma 4 model with an encoder-free design: image patches and audio waveforms are projected directly into the language model instead of passing through separate encoders. It accepts text, image, audio, and video (as frames) input and generates text, with a 256K-token context window. Google's release page dates its release to June 3, 2026, after the other Gemma 4 sizes.1498

Last reviewedEntry updated Documented release Jun 3, 2026

Availability and license

Overall availability

Public

Weights are downloadable from Hugging Face; the repositories were not gated at the time of review. Use is governed by the Apache License 2.0.1238

Availability is separate from permission: read the license before using or redistributing.

The Gemma Terms of Use cover the earlier Gemma models listed in their appendix and direct readers to the separate Gemma 4 license (Apache License 2.0) for Gemma 4 terms.65

Model-disclosure tier

Computed from the checklist below using USASI rubric v0.1. An editorial category, not a certification.
Model-disclosure tier (USASI rubric v0.1): Open-weight

The model parameters for this release can be downloaded by the public. License terms may still restrict use, redistribution, or commercial use.

How tiers are computed

Public materials checklist

Items for a model under USASI rubric v0.1. Unknown means unassessed or insufficient evidence.
Public materials checklist for Gemma 4 12B Unified
ItemStatusNotes and evidence
WeightsCan the general public download the model parameters for this release?PublicPre-trained and instruction-tuned checkpoints are published in safetensors format on Hugging Face.123
Inference codeIs code for running the model published?PublicThe model card documents inference with Hugging Face Transformers; Google DeepMind's Apache-2.0 gemma JAX library also supports Gemma 4.110
Training codeIs the code used to train the model published?UnknownThe gemma JAX library includes fine-tuning code. This catalog did not find published code used to pre-train Gemma 4.10
Training-data informationPublic = the training data itself can be obtained. Partial = composition or sources are documented without full access.PartialThe model card and technical report describe the data types (web documents, code, mathematics, images, and audio for this size) and a January 2025 cutoff; the data itself is not released.17
Training recipeAre the training configuration and procedure documented in enough detail to follow?PartialThe technical report describes the architecture, compute setup (TPUv4 and TPUv6e, JAX, Pathways), and data filtering, and says pre-training and post-training follow the Gemma 3 approach; it is not a complete recipe.7
Evaluation materialsPublic = evaluation code or prompts that let others re-run the evaluations are published. Partial = results only.PartialBenchmark results for the instruction-tuned models are reported in the model card and technical report.17

What it is useful for

The model card lists text generation, chatbots, summarization, image data extraction, NLP and vision-language research, and language-learning tools among intended uses, and documents a configurable thinking mode and native function calling. It also lists audio processing, such as speech recognition and speech translation, for this size.1

Run and use notes

Documented facts only. No hardware or performance claims are made without a cited source and stated assumptions.
  • The model card shows loading the model with Hugging Face Transformers (AutoModelForMultimodalLM) and recommends sampling with temperature 1.0, top_p 0.95, and top_k 64.1

Organization context

Other releases in the Gemma family

  • Gemma 4 26B A4BModel-disclosure tier (USASI rubric v0.1): Open-weight
  • Gemma 4 31BModel-disclosure tier (USASI rubric v0.1): Open-weight

Gemma family overview

U.S. eligibility

Project eligibility rests on documented governing or maintaining entities, not on contributors.

Eligible · basis: Documented U.S. control

The model card names Google DeepMind as the author. Google DeepMind is a research unit of Google, announced by Google's CEO in 2023, and Google's parent Alphabet Inc. has its principal executive offices in Mountain View, California, per its fiscal 2025 Form 10-K.11112

Assessed Sep 29, 2026

Sources

  1. 1.
    google/gemma-4-12B-it model card (external site: huggingface.co)

    Google DeepMind (via Hugging Face) · Model card · accessed Sep 29, 2026

  2. 2.
  3. 3.
  4. 4.
    Gemma 4 model card (external site: ai.google.dev)

    Google AI for Developers · Model card · accessed Sep 29, 2026

  5. 5.
    Gemma 4 license (Apache License 2.0) (external site: ai.google.dev)

    Google AI for Developers · License · accessed Sep 29, 2026

  6. 6.
    Gemma Terms of Use (external site: ai.google.dev)

    Google AI for Developers · License · accessed Sep 29, 2026

  7. 7.
    Gemma 4 Technical Report (external site: arxiv.org)

    Gemma Team, Google DeepMind (via arXiv) · Paper · published Jul 2026 · accessed Sep 29, 2026

  8. 8.
    Introducing Gemma 4 12B: a unified, encoder-free multimodal model (external site: blog.google)

    Google · Announcement · published Jun 3, 2026 · accessed Sep 29, 2026

  9. 9.
    Gemma releases (external site: ai.google.dev)

    Google AI for Developers · Release notes · accessed Sep 29, 2026

  10. 10.
    google-deepmind/gemma (GitHub) (external site: github.com)

    Google DeepMind · Repository · accessed Sep 29, 2026

  11. 11.
    Google DeepMind: Bringing together two world-class AI teams (external site: blog.google)

    Google · Announcement · published Apr 20, 2023 · accessed Sep 29, 2026

  12. 12.
    Alphabet Inc. Form 10-K for the fiscal year ended December 31, 2025 (external site: sec.gov)

    Alphabet Inc. (via U.S. Securities and Exchange Commission EDGAR) · Filing · published Feb 5, 2026 · accessed Sep 29, 2026

This listing is not an endorsement, a safety assessment, or a federal approval.

Support Us

Help keep USASI useful.

Optional. No USASI account required. Payment takes place on the linked provider’s website (Buy Me a Coffee).

About supporting this project