The Gemma collection of AI models openly available to Google has reached a milestone with over 150 million downloads. Google Deepmind developer relations engineer Omar Sanseviero announced the X figures over the weekend, revealing that the developer has created over 70,000 Gemmas for its AI development platform.
Google launched Gemma in February 2024, aiming to compete with other “open” models families, such as Meta’s Llama. The latest Gemma release is multimodal. This means that you can support over 100 languages using images and text. Google is also creating versions of Gemma that have been tweaked for specific applications such as Drug Discovery.
140 million downloads in about a year are a mind-boggling metric, but Gemma is probably far behind its main rival, Llama, who surpasses the 1.2 billion downloads in late April.
What’s noteworthy is that Gemma and Llama have been criticized for their habitual, non-standard licensing conditions, which say some developers are making the model a commercially risky proposal.