Google has announced a new open source artificial intelligence model Gemma, which the company claims outperforms the competition despite its smaller size.
Gemma is built from the same technical and infrastructure components as Google’s main artificial intelligence assistant, Gemini. Gemma is available in two sizes – Gemma 2B (2 billion parameters) and Gemma 7B (7 billion parameters).
Both models have pre-trained and instruction-tuned variants and are optimized to run on different AI hardware platforms such as NVIDIA GPUs and Google Cloud TPUs. Google also claims that Gemma is small enough to run on developer laptops and desktops.
According to Google, Gemma is not only the best of other open source models such as Meta Llama 2, but also significantly outperforms larger models in key benchmarks.
Google says it also used human feedback to fine-tune and strengthen the model to align it with responsible behavior and conducted assessments to reduce risks and unsafe practices. In addition, the company is releasing a responsible generative AI toolkit that includes a debugging tool, safety classification and guidance.
Source: Android Authority
