Jina Embeddings v5 Text Small by Jina AI — 1024D

677M-parameter multilingual text embedding model scoring 71.7 on MTEB English v2 and 67.7 on MMTEB. Highest performance among multilingual embedding models under 1B parameters. Built on Qwen3-0.6B-Base with distillation from Qwen3-Embedding-4B. Supports 119+ languages, 32K tokens, 1024-dim with Matryoshka truncation. Robust under truncation and binary quantization.