Hey Francesco, you're right. Calling it SOTA wouldn't be fair. But their objective was to generate a publicly available text generative model to democratize LLMs, in which BLOOM currently leads. Also, some of the languages that BLOOM generates text on were previously not having any Large language models, so in that sense it is SOTA for those languages.

For benchmarking comparison you can refer here: https://huggingface.co/bigscience/bloom#results

I will remove SOTA and replace it with something else. Thanks.