Gemma 3 12B

Gemma 3 12B is part of Google's family of lightweight open models, built from the same research and technology that underpins the Gemini line. The 12B parameter size slots into a range that Google released in more variants than earlier Gemma generations, giving it a balance of capability and efficiency for practical deployment. Google DeepMind is listed as the author on the official Hugging Face model card, which also points to the Gemma 3 Technical Report and the Responsible Generative AI Toolkit as companion documentation for anyone evaluating the model.

The model is multimodal, accepting both text and image input while producing text output, and is offered in open-weight form for both pre-trained and instruction-tuned variants. It is designed for text generation and image understanding workloads such as question answering, summarization, and reasoning, and it supports multilingual interaction across more than 140 languages within a 128K-token context window. That combination of open weights, vision input, and long-context multilingual coverage makes it a flexible fit for applications that need to mix document or image analysis with conversational or analytical text generation.