From the blog announcement >We’re releasing it initially with our lightweight model version of LaMDA. This would mean Bard is a significantly smaller model vs

From the blog announcement >We’re releasing it initially with our lightweight model version of LaMDA. This would mean Bard is a significantly smaller model vs ChatGPT, which would mean it generalizes worse for most tasks. The full featured lamda model is about 22% smaller than the GPT3 model that ChatGPT is based off of.