Gemma

Google released Gemma 4, a multimodal AI model family, with documentation, model cards, and checkpoints available on Hugging Face, alongside specialized variants including DiffusionGemma, EmbeddingGemma, FunctionGemma, MedGemma, PaliGemma 2, ShieldGemma 2, T5Gemma 2, TranslateGemma, TxGemma, VaultGemma, DataGemma, RecurrentGemma, Gemma Scope 2, Gemma-APS, Cell2Sentence-Scale, and DolphinGemma. The release includes tools for inference, fine-tuning, and deployment, as well as resources like the Gemma Cookbook and Gemma Skills for developers.

Start Here start-here Models models Inference inference Fine-Tune fine-tune Tutorials tutorials Demos and Applications demos-and-applications Gemma 4 Good Challenge gemma-4-good-challenge Gemma in Space gemma-in-space Research and Evaluation research-and-evaluation Gemma Documentation https://ai.google.dev/gemma/docs — Official documentation for selecting, running, tuning, and deploying Gemma models. Get Started with Gemma https://ai.google.dev/gemma/docs/get started — Get started running inference with the multimodal Gemma 4 models. Gemma Cookbook https://github.com/google-gemma/cookbook — Maintained notebooks, examples, workshops, and end-to-end applications. Gemma Skills https://github.com/google-gemma/gemma-skills — Reusable Agent Skills for selecting, running, and training Gemma models. Gemma Events https://luma.com/gemma-events — Overview of upcoming Gemma events. Gemma on X https://x.com/googlegemma — For news, announcements, and updates about Gemma. Gemma 4 Overview https://ai.google.dev/gemma/docs/core — An overview of the capabilities, architecture, and more of Gemma 4 models. Gemma 4 Model Card https://ai.google.dev/gemma/docs/core/model card 4 — Architecture, training, evaluation, safety, and usage details. Gemma 4 Hugging Face https://huggingface.co/collections/google/gemma-4 — Gemma 4 checkpoints + assistant models on Hugging Face. Gemma 4 QAT https://huggingface.co/collections/google/gemma-4-qat-q4-0 — Quantization-aware checkpoints for local and server inference. Gemma 4 Mobile QAT https://huggingface.co/collections/google/gemma-4-qat-mobile — Mobile-optimized checkpoints for the E2B and E4B models. Previous Gemma Model Cards https://deepmind.google/models/model-cards/ — Model cards for Gemma 1–3 and earlier Gemma families. DiffusionGemma https://huggingface.co/collections/google/diffusiongemma — Experimental discrete-diffusion text generation based on Gemma 4. EmbeddingGemma https://huggingface.co/collections/google/embeddinggemma — Compact embedding model designed for retrieval and on-device use. FunctionGemma https://huggingface.co/collections/google/functiongemma — Foundation for building specialized function-calling models. MedGemma https://huggingface.co/collections/google/medgemma-release — Models optimized for medical text and image comprehension. PaliGemma 2 https://huggingface.co/collections/google/paligemma-2-release — Vision-language models for detailed image understanding tasks. ShieldGemma 2 https://huggingface.co/collections/google/shieldgemma — Image-safety classifier built on Gemma 3. T5Gemma 2 https://huggingface.co/collections/google/t5gemma-2 — Encoder-decoder models for contextual understanding and generation. TranslateGemma https://huggingface.co/collections/google/translategemma — Translation models covering 55 languages. TxGemma https://huggingface.co/collections/google/txgemma-release — Models for therapeutic-development research. VaultGemma https://huggingface.co/google/vaultgemma-1b — Language model trained with differential privacy. DataGemma https://huggingface.co/collections/google/datagemma-release — Models and recipes for grounding responses with Data Commons. RecurrentGemma https://huggingface.co/collections/google/recurrentgemma-release — Open models based on the recurrent Griffin architecture. Gemma Scope 2 https://huggingface.co/collections/google/gemma-scope-2 — Open sparse autoencoders and interpretability tooling for studying Gemma 3. Gemma-APS https://huggingface.co/collections/google/gemma-aps-release — Abstractive proposition segmentation for decomposing text into meaningful claims. Cell2Sentence-Scale https://huggingface.co/vandijklab/C2S-Scale-Gemma-2-27B — A Gemma 2 27B model fine-tuned for single-cell biology. DolphinGemma https://deepmind.google/models/gemma/dolphingemma/ — Uses dolphin audio to help scientists study how dolphins communicate. HF Transformers https://huggingface.co/docs/transformers/model doc/gemma4 — Python library for loading, running, and fine-tuning Hugging Face models. llama.cpp https://huggingface.co/collections/ggml-org/gemma-4 — LLM inference in C/C++ with GGUF quantization. Unsloth https://unsloth.ai/docs/models/gemma-4 — Local UI to run and train LLMs and diffusion models. Ollama https://ollama.com/library/gemma4 — Get up and running with large language models locally. LM Studio https://lmstudio.ai/models/gemma-4 — Desktop application to discover, download, and run local models. vLLM https://docs.vllm.ai/projects/recipes/en/stable/Google/Gemma4.html — High-throughput and memory-efficient LLM serving engine. SGLang https://lmsysorg.mintlify.app/cookbook/autoregressive/Google/Gemma4 — Fast serving framework for large language models and vision-language models. AI Edge Gallery https://github.com/google-ai-edge/gallery — On-device ML models and examples for mobile and edge devices. LiteRT https://developers.google.com/edge/litert-lm/models/gemma-4 — Google's runtime for on-device ML deployment. JAX https://github.com/google-deepmind/gemma — Official Gemma reference implementation in JAX and Flax. React Native https://github.com/software-mansion/react-native-executorch/tree/main/apps/llm — Run on-device Gemma models within React Native using ExecuTorch. GenieX https://aihub.qualcomm.com/models/gemma 4 e4b it — Run Gemma on Qualcomm hardware. Docker https://hub.docker.com/r/ai/gemma4 — Run Gemma 4 in Docker. Gemini Enterprise Agent Platform Formerly Vertex AI https://console.cloud.google.com/agent-platform/publishers/google/model-garden/gemma4 — Fully managed enterprise AI platform on Google Cloud. OpenRouter https://openrouter.ai/google/gemma-4-31b-it — Unified API routing to multiple AI model providers. Cerebras https://inference-docs.cerebras.ai/models/gemma-4-31b — High-speed Gemma 4 inference on Cerebras. NVIDIA https://huggingface.co/nvidia/Gemma-4-31B-IT-NVFP4 — Optimized TensorRT-LLM and NVFP4 checkpoints. AMD https://www.amd.com/en/developer/resources/technical-articles/2026/day-0-support-for-gemma-4-on-amd-processors-and-gpus.html — Support for AMD ROCm GPUs and processors. AI Studio https://aistudio.google.com/app/prompts/new chat?model=gemma-4-31b-it — Web-based prototyping and development environment. Cloud Run https://docs.cloud.google.com/run/docs/run-gemma-on-cloud-run — Deploy containerized Gemma services with autoscaling GPUs. LiveKit https://livekit.com/products/inference/gemma-4 — Real-time multimodal voice and video inference infrastructure. Together AI https://www.together.ai/models/gemma-4-31b — Cloud platform for running and fine-tuning open source models. Modal https://modal.com/docs/launch/gemma-4 — Run and deploy Gemma 4 on the Modal platform. Fireworks https://fireworks.ai/models/fireworks/gemma-4-31b-it — Run and deploy Gemma 4 on the Fireworks.AI platform. BaseTen https://www.baseten.co/library/publisher/gemma/ — Run and deploy Gemma 4 on the BaseTen platform. Runpod https://docs.runpod.io/tutorials/serverless/run-gemma-7b — Experiment, train, fine-tune, and deploy Gemma. Cloudflare https://developers.cloudflare.com/workers-ai/models/gemma-4-26b-a4b-it/ — Run Gemma 4 on the Workers AI LLM Playground. Fine-Tune Gemma https://ai.google.dev/gemma/docs/tune — Official framework guide covering Keras, JAX, Hugging Face, Unsloth, Axolotl, and Google Cloud. Gemma Cookbook: Training https://github.com/google-gemma/cookbook/tree/main/docs/core — Official fine-tuning notebooks and training recipes. Tunix https://github.com/google/tunix — JAX-native library for post-training generative models. Unsloth Gemma 4 fine-tuning guide https://unsloth.ai/docs/models/gemma-4/train — Train Gemma 4 E2B, E4B, 12B, 26B A4B and 31B with Unsloth. Gemma Multimodal Tuner https://github.com/mattmireles/gemma-tuner-multimodal — Fine-tune Gemma 3n and Gemma 4 with text, images, and audio on Apple Silicon. MLX Tune https://github.com/ARahim3/mlx-tune gemma-4-audio-fine-tuning — MLX-native SFT, preference tuning, and multimodal fine-tuning with Gemma 4 support. A Visual Guide to Gemma 4 https://newsletter.maartengrootendorst.com/p/a-visual-guide-to-gemma-4 A Visual Guide to Gemma 4 12B https://newsletter.maartengrootendorst.com/p/a-visual-guide-to-gemma-4-12b A Visual Guide to DiffusionGemma https://newsletter.maartengrootendorst.com/p/a-visual-guide-to-diffusiongemma A Visual Guide to the Gemma 4 Drafters https://x.com/googlegemma/status/2051694045869879749 Variable Aspect Ratio and Variable Resolutions in Gemma 4 https://x.com/googlegemma/status/2047734622109466893 How to Use Transformers.js in a Chrome Extension https://huggingface.co/blog/transformersjs-chrome-extension How to run a local coding agent with Gemma 4 and Pi https://patloeber.com/gemma-4-pi-agent/ How to run Gemma 4 with OpenClaw https://x.com/googlegemma/status/2041512106269319328 How a Small Fix Improves Gemma 4 Vision Performance https://www.overshoot.ai/blogs/gemma-4-batched-encoder While I slept, my 5-year-old MacBook ran Gemma 4 locally and indexed a year of video https://blog.simbastack.com/indexed-a-year-of-video-locally/ Fine-tuning Gemma 4 12B on your own data https://x.com/akshay pachaar/status/2063610194618396728 Turning Gemma 4 into an Old Korean Translator https://dev.to/googleai/turning-gemma-4-into-an-old-korean-translator-hop Gemma 4 Vision Token Budget https://huggingface.co/spaces/google/gemma4 vision token budget — Explore the effect of image resolution and visual-token budgets. Concurrent Gemma https://github.com/google-gemma/cookbook/tree/main/apps/concurrent — Run and compare multiple concurrent local Gemma instances. See what 3 builders are making with Gemma 4 https://blog.google/innovation-and-ai/technology/developers-tools/gemma-4-builders/ — Various applications developed by the community. AIventure https://github.com/bebechien/AIventure — A 2D grid-based adventure game built with Phaser 3 and Angular with Gemma driving it. Gemma Chat https://github.com/ammaarreshi/gemma-chat — Local AI chat + coding agent for Apple Silicon, powered by Gemma 4 via MLX / Supports Ollama. Build with Gemma 4 and Haystack https://haystack.deepset.ai/cookbook/gemma chat rag — Runnable notebook covering RAG, visual question answering, a multimodal weather agent, and GitHub tool discovery. Gemma 4 Browser Extension https://github.com/nico-martin/gemma4-browser-extension — Local browser agent powered by Gemma 4, WebGPU, and Transformers.js. WebGemma https://github.com/NSTiwari/WebGemma — Browser playground and interactive model timeline powered by WebGPU and Transformers.js. Controlling an iOS simulator https://x.com/swmansion/status/2054949426485993669 — Gemma 4 using Argent to control an iOS simulator showcasing its capabilities in agentic workflows. Automated Video Segmentation & Tracking https://x.com/dahou yasser/status/2044372250644901899 — A demo that uses Gemma 4 + Falcon Perception for video tracking. Parking Lot Car Detection & Segmentation https://x.com/MaziyarPanahi/status/2042592050940449260 — Gemma 4 analyzes the scene, decides the questions, generates prompts, and calls SAM 3.1 as a tool. SAM 3.1 segments and returns results. Gemma 4 and MTP as a Marathon Engine https://dev.to/gde/the-local-model-that-doesnt-sleep-gemma-4-mtp-as-a-marathon-engine-4c9 — Benchmarks speculative decoding across increasing context lengths. Cactus Hybrid https://github.com/cactus-compute/cactus-hybrid — Post-trained Gemma 4 models to recognize when they are wrong, run on any framework. Damage Scout https://x.com/cerebras/status/2075671402091606416 — Damage Scout samples frames from a rental car walkaround, sends them to Gemma 4, gets back structured findings and box coordinates, then renders an annotated damage report in under 6 seconds. MedGemma Impact Challenge https://www.kaggle.com/competitions/med-gemma-impact-challenge/hackathon-winners — The winners of the MedGemma hackathon to build human-centered AI applications with MedGemma. Gemma-Translator https://github.com/google-gemma/gemma-translator — A fully offline device powered by Gemma 4 E2B built with Google Antigravity. Real-Time Voice AI with Gemma 4 https://huggingface.co/blog/cerebras-gemma4-voice-ai — Open-source cascaded voice stack using Gemma 4 for low-latency reasoning. Amazing projects https://www.kaggle.com/competitions/gemma-4-good-hackathon that harness the power of Gemma 4 to drive positive change and global impact. Trido https://www.kaggle.com/competitions/gemma-4-good-hackathon/writeups/trido — A Voice-Driven AI Whiteboard Built for the Teacher Nobody Builds For. CodeBuddy https://www.kaggle.com/competitions/gemma-4-good-hackathon/writeups/new-writeup-1778344276798 — AI Python Tutor for Indonesian Students. Port-a-Prof https://www.kaggle.com/competitions/gemma-4-good-hackathon/writeups/port-a-prof — Deeper learning, wherever you are. TriageMate https://www.kaggle.com/competitions/gemma-4-good-hackathon/writeups/triagemate-offline-first-clinical-ai-for-ghanas — Offline-first Clinical AI for Ghana's Community Health Officers. ORCA-G4 https://www.kaggle.com/competitions/gemma-4-good-hackathon/writeups/orca-g4 — On-device oral cancer intelligence for 900,000 ASHA workers in rural India. DEMENTOR https://www.kaggle.com/competitions/gemma-4-good-hackathon/writeups/dementor-edge-ai-triage-for-dementia-care — Edge AI Triage for Dementia Care. PreVillage https://www.kaggle.com/competitions/gemma-4-good-hackathon/writeups/previllage-a-govspeak-platform — A source-backed navigator for Nepal’s government services, built to find the office route, not just the form. BrailleOut https://www.kaggle.com/competitions/gemma-4-good-hackathon/writeups/new-writeup-1779137654649 — An assistive device that reads the text and images from real-world and converts it to Braille using Gemma 4 and Ollama. Gem-Care https://www.kaggle.com/competitions/gemma-4-good-hackathon/writeups/gem-care-gemma-4-good-hackathon — Gemma-4-Enriched with Multimodal Clinical-context Adaptation for Recognition Enhancement of Non-Normative Speech. Trajectix https://www.kaggle.com/competitions/gemma-4-good-hackathon/writeups/trajectix-an-agentic-flight-recorder-for-ai-infra — An Agentic Flight Recorder for AI Infrastructure Safety. TrueVoice https://www.kaggle.com/competitions/gemma-4-good-hackathon/writeups/new-writeup-1776369081947 — AI Voice Deepfake Detector. AI Conceptualizer https://www.kaggle.com/competitions/gemma-4-good-hackathon/writeups/new-writeup-1779074198441 — 3D visualizations for mechanistic interpretability and "concept spectroscopy". Acuífero·Vigía https://www.kaggle.com/competitions/gemma-4-good-hackathon/writeups/acuifero4vigia — Hybrid edge-and-citizen flood early warning for Argentina's Litoral, where every minute of warning is a life. ResQ https://www.kaggle.com/competitions/gemma-4-good-hackathon/writeups/resq — Offline Multilingual Disaster Response Coach on Gemma 4 E2B. Starcloud-1 https://www.cnbc.com/2025/12/10/nvidia-backed-starcloud-trains-first-ai-model-in-space-orbital-data-centers.html — Starcloud deployed and ran Gemma in orbit aboard an H100 GPU. NASA https://spectrum.ieee.org/nasa-ai-satellite-image-analysis — NASA runs Gemma in orbit to analyze satellite imagery and compress visual data into text for rapid, low-bandwidth disaster response. Gemma 4 Technical Report https://arxiv.org/abs/2607.02770 — The technical report covering Gemma 4 E2B, E4B, 12B, 26B A4B, and 31B. DiffusionGemma Technical Report https://arxiv.org/abs/2608.00146 — The technical report covering DiffusionGemma. Artificial Analysis https://artificialanalysis.ai/models/gemma-4-31b — Intelligence, Performance & Price Analysis. ChessBench https://x.com/googlegemma/status/2054283302090277123 — Chess LLM Benchmark Leaderboard. TERMS-Bench https://x.com/ericavaneee/status/2055868536099381638 — A benchmark for LLM negotiation agents based on economic negotiation. This is not an officially supported Google product. This project is not eligible for the Google Open Source Software Vulnerability Rewards Program https://bughunters.google.com/open-source-security .