Event banner
Azure Cognitive Search AMA: Vector search, Azure OpenAI Service, generative apps, plugins & more
Event details
- robertlee-msftJul 25, 2023
Microsoft
Adding to Liam's answer, using semantic search along with vector search is not a requirement to use vector search features itself. Customers may often find that semantic search will improve the relevance and ranking of the search results, so you may wish to perform testing on whether to use semantic search if you are more sensitive to the base cost.
If you are only using vector search, there are no additional costs to use this feature within your pre-existing search service. Please note the service limits and restrictions outlined in these pages below:
Vector limits: https://learn.microsoft.com/en-us/azure/search/search-limits-quotas-capacity#vector-index-size-limits
Vector availability and pricing: https://learn.microsoft.com/en-us/azure/search/search-limits-quotas-capacity#vector-index-size-limits
- liamca-msftJul 25, 2023
Microsoft
That is correct Brennan, unfortunately (or fortunately), the semantic model leverages GPU's that has a cost associated with it. We do understand that it makes more sense for some customers to have a more gradual ("pay per usage") type model as opposed to having the $500 cost for a large number of transactions. Please keep an eye out for that. It is important to note that for the semantic search capability, there is no cost for ingesting data as it is purely a reranking model. More details can be found here: https://learn.microsoft.com/en-us/azure/search/semantic-search-overview