Embedding Models

AWS's embedding model via Bedrock. 8K token input, 1024 dimensions. Native integration with AWS services.

BAAI's open embedding models. Top-performing on MTEB. Available in multiple sizes (small to large).

Multilingual embedding model with compression. Supports 100+ languages. Built-in int8/binary quantization.

Cross-encoder reranking model. Improves retrieval quality by reranking search results for relevance.

Late-interaction retrieval model by Stanford. Token-level matching for high-quality passage retrieval.

Microsoft's embedding models. Instruct-tuned variants for task-specific retrieval with natural language instructions.

Lightweight Python library for embedding generation. ONNX-based, runs on CPU without PyTorch dependency.

Google's latest embedding model via Vertex AI. 768 dimensions, optimized for retrieval and classification.

Open-source multilingual embeddings. 8K context, task-specific adapters (retrieval, classification, clustering).

Cross-encoder reranking model by Jina AI. Improve retrieval quality with semantic relevance scoring.

Mistral's embedding model. 1024 dimensions, 8K context. Competitive pricing via Mistral API.

High-quality embedding models with Matryoshka support. Reduce dimensions without retraining for cost savings.

Open-source embedding model with 8K context. Competitive with closed-source models. Fully auditable weights.

NVIDIA's embedding model. Top MTEB scores with strong retrieval performance across domains.

text-embedding-3-small/large models. Industry standard for text embeddings. 3072 dimensions, 8K token input.

Python framework for state-of-the-art embeddings. 5000+ pre-trained models on Hugging Face. Easy fine-tuning.

Sparse embedding model for information retrieval. Combines neural and lexical matching for hybrid search.

Specialized embedding models for code, law, finance, and general text. Top MTEB benchmark scores.

Code-specialized embedding model by Voyage AI. Best-in-class for code search and retrieval tasks.