Google's Gemma 4 12B unified multimodal model released, runs on 16GB laptop with native audio

Google introduced Gemma 4 12B, a unified encoder-free multimodal model designed to bring agentic multimodal intelligence to laptops. It is the first mid-sized model in the series to feature native audio inputs and bridges the gap between the smaller E4B and the 26B Mixture of Experts model with a reduced memory footprint. Gemma 4 models have crossed 150 million downloads. The architecture eliminates separate multimodal encoders so that vision and audio inputs flow directly into the language model backbone. For vision, a lightweight embedding module replaces the encoder using a single matrix multiplication, positional embedding and normalizations. Audio processing projects the raw signal into the text token dimensional space. The model achieves benchmark performance near the 26B version at less than half the memory footprint and runs locally on consumer laptops with 16GB of VRAM or unified memory. Gemma 4 12B is released under an Apache 2.0 license and includes Multi-Token Prediction drafters to lower latency. It supports advanced reasoning and agentic workflows. Developers can try the model in LM Studio, Ollama, the Google AI Edge Gallery App, the Google AI Edge Eloquent app and the LiteRT-LM CLI. Pre-trained and instruction-tuned weights are available on Hugging Face and Kaggle. Integration options include Hugging Face Transformers, llama.cpp, MLX, SGLang and vLLM with fine-tuning supported through Unsloth. Google released an official Skills Repository to help agents build with Gemma models. Deployment in production is available via Google Cloud services including the Gemini Enterprise Agent Platform Model Garden, Cloud Run and GKE.