Hugging Face Introduces Gemma-4, a 12B Unified AI Model
deepmind gemma google huggingface multimodal
| Source: Mastodon | Original article
Gemma4, a 12B unified AI model, emerges as a mid-tier option. It's available on Hugging Face.
Google DeepMind's Gemma 4 model family has gained a new addition, the 12B Unified model, which offers a balance between performance and resource requirements. This new model is part of the Gemma 4 family, which includes E2B, E4B, 26B-A4B, and 31B models, supporting over 140 languages and up to 256K context. As we previously reported on various AI developments, including GitHub Copilot's new pricing model and Microsoft's improved AI agent behavior control, the introduction of Gemma 4's 12B Unified model is a significant update in the field of open AI models.
The Gemma 4 models are Apache-2.0 licensed, allowing for responsible commercial use, and can be run locally on devices. This flexibility, combined with their advanced capabilities, makes them an attractive option for developers and enterprises. The 12B Unified model, in particular, provides a nice middle ground between smaller and larger models, making it an interesting choice for those looking for a balance between performance and resource usage.
As the AI landscape continues to evolve, it will be important to watch how the Gemma 4 models, including the new 12B Unified variant, are adopted and utilized by developers and organizations. With their open and flexible nature, they have the potential to drive innovation and advancement in various AI applications, from question answering and summarization to reasoning and multimodal tasks.
Sources
Back to AIPULSEN