BLOOM

BLOOM (BigScience Large Open-science Open-access Multilingual Language Model) is a 176-billion-parameter autoregressive transformer released in July 2022 by the BigScience research workshop — a year-long collaboration of over a thousand researchers coordinated through Hugging Face, with compute donated by the French government's Jean Zay supercomputer. Trained on the 366-billion-token ROOTS corpus covering 46 natural languages and 13 programming languages, it was conceived as a transparent alternative to closed models like GPT-3, with training data, code, and a full technical report published alongside the weights. It pioneered the RAIL (Responsible AI License) v1.0, which permits commercial use while banning specified harmful applications.

Limitations: BLOOM is firmly a legacy model — its 2,048-token context window and dense architecture look dated next to modern frontier models, and it has been superseded by more efficient open-weight families such as Llama and Mistral. Its few-shot performance was strong for its era but it was never instruction-tuned, so out-of-the-box chat behavior is weak. It remains valuable as a reference point for multilingual coverage and community-governed open science rather than as a deployment choice.