multilingual-e5-small

Architecture

12-layer multilingual encoder (initialized from multilingual MiniLM/E5 recipe), mean-pooled to a 384-dim embedding, trained with weakly-supervised contrastive pretraining plus supervised fine-tuning. Expects 'query:' / 'passage:' instruction prefixes.

Key Capabilities

• •384-dim multilingual embeddings (100+ languages)
• •Instruction-prefixed query/passage encoding for better retrieval
• •Small and fast for high-volume indexing
• •Strong quality-per-parameter in the E5 family

Use Cases on Mixpeek

• •Multilingual document and transcript search
• •High-volume, low-latency first-stage recall
• •Cross-lingual retrieval across an international corpus
• •Cheap recall feeding a cross-encoder reranker

Use multilingual-e5-small on Mixpeek

Build multimodal processing pipelines with this model and others. Extract features, run inference, and set up retrieval in Mixpeek Studio.

On Mixpeek, multilingual-e5-small runs as a managed extractor inside a processing pipeline. Point a bucket of sentence similarity data at it, and Mixpeek handles GPU provisioning, batching, retries, and writing the outputs into a vector store you can query.

Extractor outputs land in the Mixpeek Vector Store (MVS), where you can combine them with retrieval, reranking, and filter stages to build end-to-end search and agent-perception pipelines, no model-serving infrastructure to maintain.