Alpha
Contact
Support · Gemma 4

Will it last?

Ownership levelSubstantialnone·limited·partial·substantial·fullAnalytical input C ยท 68.8/100

This page is a projection of the one entry record, the Transparency factor that Support covers. The full verdict is set by all four factors together, floor-weighted so the weakest caps the whole.

Which domain expands which factor
  • AssessUse & modify + Transparency
  • ImplementData control + Reliability
  • UseReliability
  • SupportTransparency

Common problems & fixes

These come from the model card best practices and open repository discussions, and are not an exhaustive catalogue:

  • Odd or looping tool calls, unbalanced turn tags and lost reasoning are chat-template defects fixed in commits to the repositories (for example 'chat template - null handling, reasoning preservation, turn-tag balance, input validation' and a later response-template addition). Update to the current revision, then pin it (ev-hf-scan).
  • Thinking text in the answer or in history: parse the response with the processor's parse_response, and do not feed earlier thinking back except on tool-call turns (ev-card-31b).
  • An empty thought block with thinking off is expected for every size except E2B and E4B.
  • Weak image or audio results: put images before text and audio after text, and raise the image token budget for OCR (ev-card-31b).
  • A fine-tuned model that never stops needs the right end token: base models end with <eos>, instruction-tuned models with <turn|> (ev-tech-report).
  • Out of memory on the 26B A4B: all 25.2B parameters must be resident even though 3.8B are active (ev-docs-overview).

Versions, changelog & cadence

Gemma 4 launched on 2 April 2026 with four sizes (E2B, E4B, 26B A4B, 31B). The 12B Unified model followed (Hugging Face repository created 23 May 2026) and the model card lists five sizes (ev-blog, ev-hf-inventory). Official QAT and drafter checkpoints appeared from April to June 2026.

The technical report is v2 (v1 2 July, v2 24 July 2026) (ev-tech-report). The Hugging Face repositories were last modified between July and November 2026. The 31B-it commit history shows chat-template fixes on 10 April, 28 April, 18 May and 15 July and a response-template addition on 20 July (ev-hf-scan). The EU training-content summary is v1 dated 31 July 2026 (ev-eu-summary).

Hugging Face revisions are immutable, so the commit hash is the changelog anchor: pin it, and diff against a newer revision when upgrading.

Community & support channels

  • Hugging Face repository discussions on each google/gemma-4-* repository, which are active (ev-hf-scan).
  • GitHub issues for the google-deepmind/gemma library, and the google-gemma organisation that the model card links (ev-gemma-lib, ev-card-31b).
  • The documentation at gemma.ai (ev-docs-overview). Kaggle is named as a distribution channel but is unchecked.

There is no paid support tier from Google for the open weights. Hosted use goes through Google Cloud services named in the launch blog (ev-blog), under their own terms, which are not assessed here.

Tracked known issues

Known issues come from the repository discussions, commits and independent pages, and are partial because there is no single issue tracker for the weights:

  • Chat-template defects, including an open report dated 18 November 2026 that an image-only tool result renders its placeholder after the tool block closes (ev-hf-scan).
  • Thinking-block behaviour differs between the small and larger models (ev-card-31b).
  • A lower rank on the Artificial Analysis Intelligence Index v5.0.1 than the launch coverage suggested (ev-aa).
  • Gemma 3 derivatives and other Gemma releases keep the older Gemma Terms; check each checkpoint's licence tag before reusing a Gemma 4 conclusion (see Assess, license item).

How this scores

The ownership factor this domain covers, drawn from the one entry record.

2

TransparencyDo you know what it is: weights, training, behaviour, and legible terms?

Moderate

Weights, model card and technical report are open and a training-content summary exists. The training data is closed (the summary names no dataset), the pre-training pipeline is not released, and the summary discloses user-data and Gemini 3 synthetic-data use that cannot be inspected (ev-eu-summary, ev-tech-report).

How this scores (AOI sub-dimensions)
Provenance4/5how well we can trace and verify what went into the modelThe google organisation on Hugging Face is verified, the core checkpoints are safetensors only, each shard has an LFS sha256, the Hugging Face scan flags no files, and official quantised and drafter checkpoints are documented by use case.
Governance3/5how accountable and well-documented the publisher isAccountable, reputable publisher with a visible cadence (Gemma 1 to 4, dated chat-template and QAT updates, technical report v2) and open community discussions.
What this means for adoptionUse of these weights is owned outright: plain Apache-2.0, no gate, no use policy and no remote shut-off, which the older Gemma Terms of Use did have. What the weights are is not owned: the training data is closed, Google states that user data from its products went into training, and safety evidence is publisher-only. Adopt the instruction-tuned checkpoints behind your own guardrails and check the licence per checkpoint, because Gemma 1 to 3n and the Gemma 3 derivatives you may already run remain under the old terms.

Sources

The same evidence records as the entry sheet. Read means the text was verified; unverified means it is known to exist but not yet read.

Licenceread2026-11-19
The Gemma 4 licence page (linked as license_link by every Gemma 4 model card) is the Apache License 2.0.
Documentationread2026-11-19
The Hugging Face API (anonymous) shows the google organisation as verified.
Model cardread2026-11-19
The Gemma 4 model card (README of google/gemma-4-31B-it, same text in the other cards, plus the gemma.ai model card page) lists five sizes: E2B, E4B, 12B Unified, 26B A4B and 31B.
Documentationread2026-11-19
The Gemma 4 overview page gives a memory table per precision for each size and routes official QAT checkpoints: q4_0 GGUF for llama.cpp and LM Studio, w4a16 compressed-tensors for vLLM and SGLang, mobile builds, unquantised QAT weights and matching -assistant drafters.
Technical_reportread2026-11-19
The Gemma 4 Technical Report (arXiv 2607.02770 v2, 24 Jul 2026) states that the models are released under Apache 2.0.
Documentationread2026-11-19
Google's EU AI Act Article 53(1)(d) Public Summary of Training Content for the Gemma 4 family is v1, last updated 31 July 2026, from provider Google Ireland Limited.
Terms of serviceread2026-11-19
The Gemma Terms of Use (last modified April 1, 2026) apply only to the models in their Appendix (Gemma 1, 1.1, 2, 3, 3n and siblings; Gemma 4 is not listed) and state 'For Gemma 4 terms, see the Gemma 4 license'.
Terms of serviceread2026-11-19
The Gemma Prohibited Use Policy (last modified February 21, 2024) lists restricted uses for 'Gemma or Model Derivatives' and is bound into a contract only through section 3.2 of the Gemma Terms of Use.
Vendor announcementread2026-11-19
The launch blog (2 April 2026) says Gemma 4 is 'released under a commercially permissive Apache 2.0 license'.
Third-party analysisread2026-11-19
Hugging Face platform data for google/gemma-4-31B-it shows scansDone true with no files with issues, safetensors BF16 parameters 31,273,088,876, an LFS sha256 for each weight shard and no signature files.
Third-party analysisread2026-11-19
The Decoder (2 April 2026) reports Gemma 4 shipping under Apache 2.0 'for the first time' and says earlier Gemma versions used a more restrictive Google proprietary licence.
Third-party analysisread2026-11-19
The Artificial Analysis page for Gemma 4 31B (Reasoning) gives an Intelligence Index v5.0.1 score of 4 and 'Intelligence #14 / 15', and describes the model as at the lower end among comparable open-weights models (median 10).
Third-party analysisread2026-11-19
The European Commission page on the General-Purpose AI Code of Practice (last update 7 October 2026) lists Google among the signatories; it does not say which chapters.
Third-party analysisread2026-11-19
Ollama library page for gemma4: tags e2b, e4b, 12b, 26b and 31b; capability tags vision, tools, thinking and audio; 29.5M downloads; run command 'ollama run gemma4'.
Documentationread2026-11-19
google-deepmind/gemma README and LICENSE: a JAX library to use and fine-tune Gemma with Gemma 4 checkpoint paths; code under Apache-2.0; the library is not the pre-training pipeline; issues go to the repository's GitHub tracker.
Securityread2026-11-19
Google's organisation-wide GitHub SECURITY.md says to report vulnerabilities at g.co/vulnz and that the Google Security Team processes a report within a day and responds within a week.
Third-party analysisunverified2026-11-19
The ADAPT Centre AI Accountability Lab evaluation of the Gemma 4 public training summary is referenced by a secondary blog.