IBM Granite Embedding Multilingual R2 supports 200+ languages and 32K context
IBM released 311M and 97M text-embedding models under Apache 2.0. They support code and more than 200 languages and extend context to 32,768 tokens, 64× the previous generation. The 97M version scored 60.3 on MTEB v2 retrieval, leading the reported open multilingual models under 100M parameters. It works with LangChain, LlamaIndex, and Haystack.
Visit original source