Model selection wizard

Embedding Model Decision Tree

Interactive wizard that helps enterprises select the optimal embedding model based on language support, domain specificity, and budget constraints. Tailored for RAG & Knowledge workflows focusing on embedding models.

Selecting the right embedding model is critical for retrieval-augmented generation (RAG) and knowledge management applications. Different models vary significantly on supported languages, domain adaptability, and cost structures.

This interactive decision tree guides you through key considerations including your target language, whether you require a domain-specific model, and your budget range. The outcome recommends embedding models aligned with your enterprise priorities.

Inputs

Primary language for embeddings

Choose the language your documents or queries primarily use.

Require domain-specific embeddings?

YesNo, general-purpose is fine

Domain-specific models can improve relevance for specialized content (e.g., legal, medical).

Monthly embedding budget ($ USD)$

Approximate monthly spend available for embedding API or service costs.

Estimated monthly embedding calls

Number of embeddings generated per month is a key determinant of cost.

Result

Recommended embedding model(s)

if (primaryLanguage == 'english' && domainSpecific == 'no' && monthlyBudget >= 500 && embeddingVolume != 'high') return 'OpenAI text-embedding-3-small v3'; else if (primaryLanguage == 'english' && domainSpecific == 'yes' && monthlyBudget >= 1500) return 'Cohere specialized domain embeddings'; else if (primaryLanguage == 'multilingual' && monthlyBudget >= 1200) return 'Google Vertex Multilingual Embedding'; else if (embeddingVolume == 'high' && monthlyBudget < 600) return 'Open Source models like SBERT'; else return 'Contact vendor sales for custom solution';

—

Your embedding model recommendation

Based on your inputs, a cost-effective and performant model fit for your language and domain needs has been identified.

Note

Embedding model pricing and capabilities vary by vendor. OpenAI's text-embedding-3-small v3 starts at $0.0004 per 1k tokens (OpenAI pricing as of 2024 Q2). Cohere and Google Vertex embed specialized high-cost models ranging $2,000–$3,000 monthly minimum. Open source alternatives like SBERT require infrastructure management but incur no per-call fees.

Subsequent sections unlock after submit