What are the new Gemma 4 models from Google?

The Gemma 4 family consists of open-weight models with varying parameter counts, designed for both edge devices and powerful machines, offering advanced AI capabilities.

Where can developers find the Gemma 4 model weights?

The model weights for the Gemma 4 family are available for download through platforms like Hugging Face, Kaggle, and Ollama.

Home / Technology / Google Unleashes Open-Source AI Powerhouse

Google Unleashes Open-Source AI Powerhouse

Q: How intelligent are the Gemma 4 models per parameter?

Google claims the Gemma 4 models demonstrate an "unprecedented level of intelligence-per-parameter," with some variants outperforming larger models.

2 Apr

•

Summary

Google released its Gemma 4 family of open-weight models.
Models offer varied parameter counts for different devices.
Gemma 4 models boast high intelligence-per-parameter performance.

Google Unleashes Open-Source AI Powerhouse

Google has released its new Gemma 4 family of open-weight models, extending the advanced technology behind its proprietary Gemini models to the open-source community. This new family includes four versions: two smaller models with 2 and 4 billion parameters for edge devices, and two larger models with 26 billion (Mixture of Experts) and 31 billion (Dense) parameters for more powerful hardware.

Google highlights Gemma 4's efficiency, claiming an "unprecedented level of intelligence-per-parameter." Notably, the 31-billion and 26-billion parameter models achieved impressive rankings on Arena AI's text leaderboard, surpassing models significantly larger than themselves.

These versatile models can process video and images, making them suitable for tasks like optical character recognition. The smaller versions also handle audio inputs and speech. Furthermore, Gemma 4 models can generate code offline and support over 140 languages. Released under an Apache 2.0 license, they offer developers greater flexibility and control.

Model weights for the Gemma 4 family are accessible through Hugging Face, Kaggle, and Ollama, encouraging broad adoption and development.

Disclaimer: This story has been auto-aggregated and auto-summarised by a computer program. This story has not been edited or created by the Feedzop team.

Google Unleashes Open-Source AI Powerhouse

2 Apr

•

Summary

Google released its Gemma 4 family of open-weight models.
Models offer varied parameter counts for different devices.
Gemma 4 models boast high intelligence-per-parameter performance.

Model weights for the Gemma 4 family are accessible through Hugging Face, Kaggle, and Ollama, encouraging broad adoption and development.

Disclaimer: This story has been auto-aggregated and auto-summarised by a computer program. This story has not been edited or created by the Feedzop team.