Gemma AI Models Surpass 150 Million Downloads - Google AI

Gemma AI Model Downloads Surpass 150 Million

A significant achievement has been reached with Google’s Gemma collection of open-source AI models, now totaling over 150 million downloads. This milestone was recently publicized by Omar Sanseviero, a developer relations engineer from Google DeepMind, via X (formerly Twitter).

Developer Activity and Model Variants

Sanseviero also shared that the developer community has been highly active, generating more than 70,000 distinct variations of Gemma on the Hugging Face AI development platform.

Gemma's Capabilities and Launch

Initially released in February 2024, Gemma was designed to provide a competitive alternative to other open-source model families, such as Meta’s Llama.

Recent Gemma releases are now multimodal, enabling them to process both image and text data. Furthermore, the models offer support for over 100 different languages.

Specialized Gemma Versions

Google has also developed specialized versions of Gemma, which have been fine-tuned for specific use cases. An example of this is the creation of models tailored for applications in the field of drug discovery.

Comparison to Llama and Licensing Concerns

Although 150 million downloads within approximately a year represents a substantial figure, Gemma currently lags behind its primary competitor, Llama. Llama exceeded 1.2 billion downloads by late April.

It is important to note that both Gemma and Llama have faced scrutiny regarding their licensing terms. Some developers have expressed concerns that these terms are unconventional and could pose risks when utilizing the models for commercial purposes.