XSearch Model Options
The XSearch Models page is where you manage the model and training-level configurations related to XSearch.
Impact on XSearch Engines
Changes made here will impact all engines that use these models.
There are three models that are publicly available in our library, each with their own configurations:
.png)
Consult our sales and customer service teams to determine which model is right for you, or if we need to create a new model for your needs.
FAQ
What is the difference between GenSearch, Hybrid Search, and DeepSearch?
GenSearch is image-focused, meaning that much of the information from the product catalog might not be useful or searchable when using GenSearch Models in your engine.
DeepSearch is a neural search, which is an explicit (keyword) search combined with a neural network, that use product attributes from your catalog to make inferences.
Hybrid Search combines keyword-based and vector-based (semantic) search with an image-focused search to balance performance and relevance.
When should I use each model?
Use GenSearch when you want relevance fine-tuned with natural language understanding and a fast response time.
Use Hybrid Search for balanced performance in cases where both keyword match and semantics matter (e.g., high-volume catalog sites).
Use DeepSearch when recall is a priority—especially for non-keyword-friendly catalogs or broad, vague user queries.
How are XSearch Engines connected to XSearch Models?
XSearch models are used by XSearch engines to determine how to process, rank, and return search results based on user queries. The engine uses the logic and capabilities of that model to:
Interpret the query (keyword or semantic parsing)
Retrieve results (exact match vs. semantic embedding vs. deep recall)
Apply boosts and ranking (e.g., personalization, business rules)
Return final search results to the user