ellamind/propella-1-0.6b

The ellamind/propella-1-0.6b is a 0.6 billion parameter multilingual large language model from the propella-1 family, developed by ellamind. It is specifically designed for annotating text documents across 18 properties in six categories, including content quality, safety, and geographic relevance. This model excels at high-throughput, accurate data curation for LLM training datasets, supporting 57 languages and handling various text formats like web pages, PDFs, and code. Its small size and fp8 training enable fast inference for scalable data processing.