Turbopuffer review

Entry pricing starts at $16. The group median for the same measure is €13.15, with the middle of the field sitting above €4.38 and below €21.38. Turbopuffer therefore starts above the median and below the upper quartile.

Turbopuffer lists a single engine, tpuf, and no other engine appears in the range. Teams that need relational queries or an additional store alongside vector search should plan for a different service rather than an extra engine here.

Vectors are supported up to 10,752 dimensions. That ceiling covers the wider embedding models, and anything beyond it needs to be reduced or handled elsewhere, so check the limit against your embedding model before you build the schema.

Turbopuffer lists no automated backups as its backup position. That means restore points are not created on a schedule by the service, so teams that need scheduled recovery should keep their own copy of the data elsewhere and check the restore path before relying on it.

Support runs on email, help centre and community only. Responses arrive in writing, and the help centre and community channel cover the routine questions, so a team that needs live voice support for an incident will not find it here.

It fits teams that need vector retrieval and want a narrow, focused service, rather than a general-purpose database. It was founded in 2023 and is operated by turbopuffer Inc., so short-lived experiments and production retrieval workloads sit with the same vendor.

Turbopuffer starts at $16, while UCloud starts at $0.96 and Vultr at $2.50. The gaps are modest in absolute terms, so for a small team the engine choice and the query surface matter more than the entry price.

On price from, Turbopuffer holds position 42 of the 70 brands ranked. That is a mid-table spot, and the cost score of 5.5 fits a price that is neither cheap nor expensive for the group.