Gemma 4 12B

Evidence (7)

• 5Hacker News·7/20/2026product_launch
  Nativ lets Mac users run open models locally

  Nativ is a curated library for selecting and running open AI models locally on Mac hardware. It recommends partner models based on the user's machine and highlights options from Google, Cohere, and Liquid AI with different sizes and context windows.

• 5Hacker News·7/15/2026open_source
  Gemma 4 runs on a 13-year-old Xeon with CPU-only optimizations

  A Hacker News post describes running Google's Gemma 4 26B mixture-of-experts model on a repurposed 13-year-old HP server with dual Ivy Bridge Xeons and no GPU. After adjusting the inference code to handle pre-AVX2 CPUs, the model reaches about 5.2 tokens per second on the basement machine.

• 7Hugging Face Blog·7/1/2026product_launch
  Hugging Face and Cerebras enable real-time Gemma 4 voice AI

  Hugging Face and Cerebras showcased a real-time speech-to-speech demo built around Gemma 4 31B, aiming to reduce latency in voice AI. The open modular pipeline combines Parakeet for speech recognition, Cerebras for fast model inference, and Qwen3TTS for text-to-speech, and is already used in Reachy Mini robots.

• 6Hugging Face Blog·6/9/2026model_release
  Google DeepMind introduces Gemma 4 12B multimodal model

  Google DeepMind has introduced Gemma 4 12B, a unified multimodal model designed without a separate encoder. The model is positioned as a more integrated approach for handling multiple input types in one architecture.

• 6Hugging Face Blog·5/27/2026product_launch
  Reachy Mini can now run fully locally

  Hugging Face announced that Reachy Mini’s conversation stack can now run entirely on local hardware instead of sending audio to a server. The setup uses a cascaded speech-to-speech pipeline with llama.cpp, Silero VAD, Parakeet-TDT 0.6B v3 for STT, and Qwen3-TTS, enabling private offline conversations with the robot.

• 6Hugging Face Blog·4/23/2026open_source
  Using Transformers.js in a Chrome Extension

  Hugging Face published a guide on building a Chrome extension that runs local AI features with Transformers.js under Manifest V3. The article outlines an architecture where a background service worker hosts the model, a side panel provides chat UI, and a content script handles page extraction and highlighting. It also shares practical notes on runtime constraints, model loading, and messaging between extension contexts.

• 8Google DeepMind·4/22/2026research
  Google DeepMind unveils Decoupled DiLoCo for resilient distributed AI training

  Google DeepMind introduced Decoupled DiLoCo, a distributed training architecture that splits large model training across asynchronous compute islands with lower bandwidth needs. The approach is designed to keep training running through hardware failures and data-center disruptions, while matching conventional training performance in Gemma 4 experiments.