BLOOM

BLOOM is a large decoder-only Transformer trained by the BigScience collaboration with an emphasis on openness, multilingual coverage, and reproducibility. The model generates and rewrites text, translates between languages, answers questions, and supports light coding tasks, and it can be adapted with instruction or domain fine-tuning for assistants and RAG workflows. Its release included model checkpoints, training details, and an open license with usage guidelines, which made BLOOM a common baseline for academic studies and production experiments that value transparency, self hosting, and multilingual support. While newer families often provide deeper reasoning and longer contexts, BLOOM remains a solid open foundation that is easy to analyze, extend, and integrate into custom pipelines.