Google has launched a new artificial intelligence (AI) large language model, Gemma, which is free to download and can run on users' computers.
The AI is optimised for Nvidia graphics processing units (GPUs) for performance, and can also run in the cloud.
Gemma is based on the technology for its commercial Gemini generative AI, and can be had with two billion or seven billion parameters, which are the neural network values used for model behaviour.
While the model is open access, it is not fully open source however.
The number of parameters is considered lightweight, but Google said that the seven billion version of Gemma still outperforms Meta's Llama-2 AI.
Llama-2 does, however, have a 70 billion parameter model that Google didn't benchmark Gemma against.
Google said it has incorporated safety measures for Gemma, based on its AI Principles and responsible design.
This included filtering out sensitive information from the data sets used for Gemma, Google said.
"As part of making Gemma pre-trained models safe and reliable, we used automated techniques to filter out certain personal information and other sensitive data from training sets," the Gemma team said.
We welcome your comments below. If you are not already registered, please register to comment.
Remember we welcome robust, respectful and insightful debate. We don't welcome abusive or defamatory comments and will de-register those repeatedly making such comments. Our current comment policy is here.