Skip to the content
Know your tech before you buy it. Makers: get listed from $49/mo

Vespa

vespa.ai

  • Free plan
  • Open source
  • Windows
  • macOS
  • Linux
  • Web
  • API
  • Self-hosted
8.7 Club score Visit vespa.ai

What Vespa is

Vespa is an AI search platform for online applications that work with changing data. It combines retrieval, ranking, machine-learning inference, and real-time serving. Search options include vectors and tensors, positional text, and structured data; ranking signals can be combined using tensor functions or models such as ONNX and XGBoost. Vespa distributes functions across node clusters and allows cluster size to change without affecting queries or writes. Teams can self-manage deployments, use the Vespa Kubernetes Operator, or deploy on Vespa Cloud. The Cloud service handles encryption at rest and in transit, uses node and endpoint certificates, and updates operating system patches automatically. Its Enclave option runs applications in a customer's AWS, Azure, or GCP account, with those providers' resource costs charged separately. A Vespa Cloud trial comes with $300 in usage credits and requires no credit card; the application stops when the credits are used. Cloud plans have quotas, and Startup is limited to development zones. Local deployment guidance lists Linux, macOS, and Windows 10 Pro.

Who it suits

Vespa is for teams building AI applications where search quality, ranking, fresh real-time data, or growing retrieval traffic matters. It offers both self-managed and cloud deployment routes.

Good to know, and what to watch

  • Supports vector, tensor, positional text, and structured search
  • Ranking can use ONNX and XGBoost models
  • Cluster size can change without interrupting queries or writes
  • Supports self-managed and cloud deployments
  • Trial includes $300 in usage credits
  • No free plan
  • Cloud plans have quotas
  • Startup is limited to development zones
  • Startup support is community-only with no SLA

Startup

—

Basic

—

Commercial

—

Enterprise

—

Getting started

  1. Open the Vespa website
  2. Choose self-managed deployment, the Vespa Kubernetes Operator, or Vespa Cloud
  3. For local use, prepare Linux, macOS, or Windows 10 Pro with Docker Desktop or Podman Desktop
  4. Select a Vespa Cloud plan or start the free trial
  5. Use the trial’s $300 usage credits without entering a credit card
  6. Connect through available APIs or Python and Java SDKs

Limits to check

The Vespa Cloud trial stops the application when its $300 usage credits run out. Every Vespa Cloud plan has a quota, and Startup is restricted to development zones with no SSO or autoscaling.

The Club's review

Vespa is a search and serving platform for teams building AI applications around changing data. It suits teams that need to tune retrieval and ranking together; for a simpler vector lookup, its breadth and deployment choices may be more than the job requires.

Overview

Vespa combines vector and tensor search, positional text search, and structured-data search with ranking, model inference, and real-time serving. That range is useful when an application depends on relevance as well as retrieval: teams can use hybrid search, sparse vectors, metadata filtering, and HNSW indexes, then combine ranking signals rather than treating search as a single lookup step. Python and Java SDKs are supported.

Deployment can be self-managed, run with the Vespa Kubernetes Operator, or hosted on Vespa Cloud. This flexibility lets teams choose how much infrastructure to operate, but also puts more weight on selecting a deployment model and understanding cloud quotas.

Key features

Search, ranking, and inference

Ranking signals can be combined with tensor functions or machine-learned models in ONNX and XGBoost formats, and ONNX Runtime supports model inference. This is a strong fit for applications where ranking logic and search need to work closely together. Teams that only need basic vector retrieval may not benefit from the added range.

Scaling and operations

Vespa distributes functions across node clusters and can change cluster size without affecting queries or writes. That makes it relevant to teams expecting retrieval traffic to grow while their application remains active. Self-managed users can run it on Linux, macOS, or Windows 10 Pro on x86_64 or arm64, using Docker Desktop or Podman Desktop.

Cloud security and integrations

Vespa Cloud encrypts data at rest and in transit, uses node and endpoint certificates, and applies OS patches automatically. Vespa Cloud Enclave can run applications inside a tenant’s AWS, Azure, or GCP account, but provider resource charges come on top of Vespa Cloud costs. Its client for OpenAI-compatible APIs, including OpenAI, Google Gemini, Anthropic, Cohere, and Together.ai, is marked beta, so teams should treat it accordingly.

Pricing

Vespa Cloud plans have quotas. The cloud free trial provides $300 in usage credits without requiring a credit card; the application stops when those credits run out. Vespa has no free plan, so the trial is a limited way to evaluate the service rather than a continuing no-cost tier.

  • Startup — $0.05 a month: billed resource rates are $ 0.05 / hour per vCPU, $ 0.005 / hour per GB of memory, $ 0.0002 / hour per GB of disk, and $ 0.03 / hour per GB of GPU memory. Shared resources, development zones only, community support with no SLA, and no SSO or autoscaling make this a constrained starting point for development, not a production-support choice.
  • Basic — $0.10 a month: initial hourly rates are $ 0.1 / hour per vCPU, $ 0.01 / hour per GB of memory, $ 0.0004 / hour per GB of disk, and $ 0.07 / hour per GB of GPU memory; rates go down with volume. It suits applications that do not need 24/7 operational support.
  • Commercial — $0.15 a month: initial hourly rates are $ 0.145 / hour per vCPU, $ 0.0145 / hour per GB of memory, $ 0.0005 / hour per GB of disk, and $ 0.1 / hour per GB of GPU memory; rates go down with volume. It adds 24/7 operational support, backup, and disaster recovery, making it the clearer fit when continuity and round-the-clock help matter.
  • Enterprise — $0.18 a month: initial rates are $ 0.18 / hour per vCPU, $ 0.018 / hour per GB of memory, $ 0.0007 / hour per GB of disk, and $ 0.125 / hour per GB of GPU memory, with a $20,000 minimum monthly spend. That minimum makes this a substantial commitment, not a casual upgrade.
  • Enterprise * — $0.18 a month: initial hourly rates match Enterprise, with prices going down with volume and a $20,000 minimum monthly spend. It includes backup and disaster recovery, with SSO and dedicated support options.
  • Self Managed — custom pricing: includes self-managed deployment and unlimited support cases, with a dedicated support representative available. This fits teams that want to operate Vespa themselves while retaining a support path.

For Startup, the low entry rate comes with development-only zones and community support; Basic also lacks 24/7 operational support. Commercial is the first tier described with that support, while the Enterprise minimum is likely to rule it out for teams without a substantial monthly budget.

Platforms

Vespa supports API, Linux, macOS, self-hosted, web, and Windows deployments. Local installation supports Linux, macOS, and Windows 10 Pro on x86_64 or arm64 with Docker Desktop or Podman Desktop.

Who it's for

Vespa is best suited to teams building AI applications where search quality, ranking, fresh real-time data, or growing retrieval traffic are important. Its model and ranking options serve relevance-sensitive workloads, while cluster scaling and deployment choices suit teams prepared to make infrastructure decisions. A team seeking a no-cost hosted plan or a narrowly focused vector lookup should consider whether Vespa’s operational and plan complexity is justified.

Pros and cons

  • Pro: Multiple search modes and configurable ranking support applications that need more than a basic vector match.
  • Pro: Cluster size can change without affecting queries or writes, a practical advantage for changing workloads.
  • Pro: Self-managed, Kubernetes Operator, and cloud deployments give teams a choice of operational model.
  • Con: Cloud quotas apply across plans, and Startup is restricted to development zones with no SLA.
  • Con: No free plan means ongoing use requires a paid plan after the trial credits are exhausted.
  • Con: The $20,000 minimum monthly spend makes Enterprise a high-commitment option.
  • Con: Enclave customers pay cloud-provider resource costs in addition to Vespa Cloud charges.

Alternatives

Pinecone is worth considering if a freemium hosted option with a free plan and a defined Starter allowance—up to 5 indexes, 100 namespaces per index, 2 GB storage, 2M monthly write units, and 1M monthly read units—better suits the workload.

Chroma is an alternative for teams seeking a freemium option with broad platform support and an Apache 2.0 open-source codebase also used by Chroma Cloud.

Upstash Vector has a free tier capped at 10K daily queries or updates, 200M vectors × dimensions, 1,536 maximum dimensions, 100 namespaces, and 1 GB of data and metadata; it may fit a workload that stays within those limits.

Zilliz Cloud offers a free cluster with 5 GB storage, 2.5M vCUs per month, and up to 5 collections, making it an option when those free-cluster limits are sufficient.

KDB.AI offers a free Cloud Starter Edition with 4 GB memory per instance, 30 GB storage, and a 10 MB query-size limit.

Qdrant offers a free-forever single-node cloud cluster with 0.5 vCPU, 1GB RAM, and 4 GB disk, which may be preferable for a small workload that fits those caps.

Weaviate provides a free tier with one cluster, 100,000 objects, 1 GB memory, 10 GB disk, one collection, and up to three tenants.

Actian Zen is another alternative.

Browse the Vector Databases, Search Databases, and Database Software directories for more options.

Verdict

Choose Vespa if your AI application needs retrieval, ranking, and real-time serving to operate as one system, especially as data or traffic changes. Its range of search methods, model support, and deployment choices are compelling for teams with specific relevance needs. Look elsewhere if a free ongoing plan, straightforward vector retrieval, or a low-commitment enterprise tier matters more than that control.

The verdict

Vespa combines search, ranking, inference, and real-time serving, with deployment options for different operating models. Review the cloud quota and support limits, especially for the Startup plan.

Questions people ask

Does Vespa have a free plan?

No. Vespa is paid, starting from $0.05, and it offers a free Vespa Cloud trial with $300 in usage credits.

Which platforms can run Vespa?

Vespa lists API, Linux, macOS, self-hosted, web, and Windows platforms. Local guidance covers Linux, macOS, and Windows 10 Pro on x86_64 or arm64.

Can Vespa be self-hosted?

Yes. It can be self-managed, operated with the Vespa Kubernetes Operator, or deployed on Vespa Cloud.

What models does Vespa support for ranking?

Ranking supports ONNX and XGBoost models, with inference through ONNX Runtime.

Does Vespa connect to generative AI APIs?

A beta client supports OpenAI-compatible APIs, including OpenAI, Google Gemini, Anthropic, Cohere, and Together.ai.

What the maker publishes

API
Applications can write data through HTTP/2, the Document API, or client tools, and can query through the Query API or custom request handlers. source
Cloud service trial
The Vespa Cloud page offers a free trial without requiring a credit card. source
Company legal name
The Norwegian business registry lists the company as VESPA.AI AS. source
Company size
Gartner Peer Insights lists the company as having 11–50 employees. source
Customer cloud accounts
Vespa Cloud Enclave runs applications inside a tenant's own AWS, Azure, or GCP account, with cloud-provider resource costs charged in addition to Vespa Cloud costs. source
Deployment
Vespa can be self-managed, run with the Vespa Kubernetes Operator, or deployed on Vespa Cloud. source
Documentation depth
Vespa documentation is organized into guides, tutorials, and reference material covering features and APIs. source
Extensible processing
Application packages can include custom Java components to process queries, results, and document writes. source
Founded
Gartner Peer Insights lists Vespa’s founding year as 2023. source
Free plan
No source
Funding
Vespa.ai announced a $31 million Series A investment from Blossom Capital on November 1, 2023. source
Grouping and aggregation
Search results can be grouped into nested groups and aggregated for uses such as navigation and result diversity. source
Headquarters
Vespa says its headquarters and engineering hub are in Trondheim, Norway. source
Integrations
Vespa provides a client for OpenAI-compatible APIs, including OpenAI, Google Gemini, Anthropic, Cohere, and Together.ai; this feature is marked beta. source
Integrations
Vespa Cloud Enclave can run applications in customer accounts on AWS, Azure, or Google Cloud Platform. source
Licence
The Vespa repository states that its code is licensed under Apache 2.0. source
Limit: Cost and resources
A PeerSpot reviewer said Vespa was costly relative to Elasticsearch and resource-intensive, and advised that its learning curve and infrastructure costs may make it a poor fit outside enterprise use. source
Limit: Documentation and setup
A PeerSpot review summary identifies documentation for beginners and setup difficulty as challenges. source
Limit: Learning curve
A PeerSpot reviewer cited a steep learning curve as a challenge when using Vespa for e-commerce search. source
Limit: Migration and authentication
A PeerSpot review summary says migration scripts are difficult and mentions a lack of user authentication. source
Machine-learned ranking
Vespa supports ranking expressions using signals from queries and documents, including models in ONNX and XGBoost formats. source
ML models
Vespa supports ONNX and XGBoost models for ranking, and model inference using ONNX Runtime. source
Notable limit
All Vespa Cloud plans have a quota, and the Startup plan is limited to development zones. source
Partners
Vespa lists OpenAI, NVIDIA, Anyscale, and LangChain among its industry partners. source

Checked against the maker's pages on 14 Oct 2026.