Project AURIONAI Use-Cases Research & Innovation
edge LLMfamily

Gemma 3

Google Gemma 3 family; small sizes in Ollama and widely used bases on Hugging Face.

Google
Levelfamily
Classedge LLM
Parametersnot in the notes
Contextnot in the notes
Released12 Mar 2025
Opennessunknown
Runs onnot in the notes
Statusactive

Benchmark Results

vendor-reportedmaintainer-run
MedQAdomain
Gemma 3 4BBaseline in MedGemma reportvendor-reported
50.7
050100

Score in %. Higher is better.

Jul 2025 arxiv.org
SEA-HELMSoutheast Asian and Indonesian
gemma-3-27b-itCheckpoint gemma-3-27b-it as listed; release view (data last updated 18 Sep 2026); score is the average of 2000 bootstrapsmaintainer-run
70.41 95% CI [-1.47, +1.49]
SEA-HELM leaderboard, 58 models, Indonesian (ID)18 Sep 2026 leaderboard.sea-lion.ai
gemma-3-27b-itCheckpoint gemma-3-27b-it as listed; release view (data last updated 18 Sep 2026); score is the average of 2000 bootstrapsmaintainer-run
64.54 95% CI [-0.74, +0.72]
SEA-HELM leaderboard, 58 models, SEA average18 Sep 2026 leaderboard.sea-lion.ai
Gemma-3-4B-itCheckpoint gemma-3-4b-it as listed; release view (data last updated 18 Sep 2026); score is the average of 2000 bootstrapsmaintainer-run
57.71 95% CI [-1.76, +1.67]
SEA-HELM leaderboard, 58 models, Indonesian (ID)18 Sep 2026 leaderboard.sea-lion.ai
Gemma-3-4B-itCheckpoint gemma-3-4b-it as listed; release view (data last updated 18 Sep 2026); score is the average of 2000 bootstrapsmaintainer-run
47.84 95% CI [-0.75, +0.74]
04080

Score in score. Higher is better.

SEA-HELM leaderboard, 58 models, SEA average18 Sep 2026 leaderboard.sea-lion.ai
Where this comes from
Every score names who ran it: vendor-reported, maintainer-run or independent. A rank is shown with the population it was ranked in (gap rule 3).

Lineage

finetune
Gemma 3 to ShieldGemma 2 4B: finetune, medium confidenceGemma 3Gemma 3Google, Mar 2025ShieldGemma 2 4BShieldGemma 2 4BGoogle, Apr 2025
20252026Gemma 3 4BGemma 3 4B, Mar 2025gemma-3-27b-it, Mar 2025Gemma-3-4B-it, Mar 2025Gemma 3 270MGemma 3 270M, Aug 2025

2 members have no release date in the notes, so they are not on this timeline.

GoogleShieldGemma 2 4B is a child (finetune)medium confidencenotes
GoogleGemma 3 270M belongs to this familyedge LLM5 Aug 2025
GoogleGemma 3 4B belongs to this familymultimodal LLM12 Mar 20254B
Googlegemma-3-27b-it belongs to this familyLLM12 Mar 2025
GoogleGemma-3-4B-it belongs to this familyLLM12 Mar 20254B
GoogleGemma 3 1B belongs to this familyedge LLM
GoogleGemma 3 27B belongs to this familymultimodal LLM27B
Where this comes from
Parent and child relations from model cards, Hugging Face base_model fields and the genealogy tables of the research notes. The tree shows two generations each way, 2 models. Confidence and evidence are the notes' own.

R&D Topics

Identity

Developer
Google
Family
not in the notes
Released
12 Mar 2025 sourcesannouncement page
Weights published
–
Record checked
10 Oct 2026
Where this comes from
Developer and dates from the research notes. Release date is the vendor's public announcement; weights date is when open weights were published.

Architecture and Size

Architecture
not in the notes
Total parameters
– sources
Active parameters
–
Checkpoint parameters
–
Context length
not in the notes
Input
text
Output
text
Where this comes from
Parameter counts as the developer states them. Where the checkpoint holds a different total, both are shown (register A16).

Openness and Access

Openness
unknown
Licence
Gemma Terms of Use sources
Access routes
not in the notes
Where this comes from
The licence file governs over registry tags and aggregator labels.

Deployment

Runs on
not in the notes
Footprint
not in the notes
Runtimes
not in the notes
Where this comes from
Deployment tiers and runtimes as named in the notes. Footprint is the notes' figure, not a measurement by AURION.

Indonesia

Indonesian support
not stated
Regional languages
not in the notes
Jakarta hosting
unknown
Languages
not in the notes
Where this comes from
Indonesia fields are carried on every model; unknown is an allowed value.

Industry Use Cases

No industry use case links this model.

Sources

  1. google (Google) huggingface.co https://huggingface.co/google/models?search=shieldgemma
  2. library ollama.com https://ollama.com/library
  3. meta-llama/Llama-3.1-8B-Instruct , Hugging Face huggingface.co https://huggingface.co/meta-llama/Llama-3.1-8B-Instruct
  4. MedGemma Technical Report arxiv.org https://arxiv.org/html/2507.05201v2
  5. google/medgemma-27b-text-it , Hugging Face huggingface.co https://huggingface.co/google/medgemma-27b-text-it
  6. blog.google/technology/developers/gemma-3 https://blog.google/technology/developers/gemma-3/
  7. huggingface.co/google/gemma-3-27b-it https://huggingface.co/google/gemma-3-27b-it

To cite this page: Project AURION, AI Use-Cases Research & Innovation, "Gemma 3", https://aurion.semai.id/models/google-gemma-3/, records of October 2026.