Embedding / RAG
#1
84
BGE M3
BAAIFlagship multilingual BGE checkpoints for dense, sparse, and hybrid retrieval setups.
Best for
Open-source RAG, local vector DBs, and academic baselines.
Overview
Late-interaction friendly embeddings and small specialized models.
Strengths & direction
Multimodal retrieval experiments and API-first prototypes. Optimized for product teams evaluating practical fit—not lab benchmarks alone.
Pricing
Freemium / API
Verify on the vendor site before production commitments.
Capabilities
embeddingsRAGmultimodal
You might also compare
Related models
Paired by category proximity and similar usefulness scores.
Embedding / RAG
#2
88
Cohere Embed v4
CohereLatest multilingual Cohere embed family for retrieval, rerank, and search stacks.
Best for
Search ranking, classification features, and Cohere-first stacks.
Embedding / RAG
#3
90
Gemini embedding 004
GoogleShipping Gemini-class embedding endpoint for Vertex AI and Gemini API retrieval stacks.
Best for
GCP-native RAG, multimodal-ish retrieval stacks, and batch embedding.
Embedding / RAG
#4
92
text-embedding-4-large
OpenAIOpenAI’s latest large embedding model tuned for retrieval, deduping, and RAG backends.
Best for
Enterprise RAG, semantic search, and hybrid vector indexes.
Image Generation
#5
82
Recraft
RecraftDesign-native generator for brand-consistent graphics and vector assets.
Best for
Brand kits, UI illustrations, and iterative design systems.