Buying snapshot
Scroll sideways to see all columns. With a keyboard, focus this table and use the arrow keys.
| Decision point | What Zilliz states |
|---|---|
Deployment | Fully managed Serverless or Dedicated clusters, or opt-in Bring Your Own Cloud (BYOC), which has the same features and experience as SaaS Dedicated clusters |
Cloud providers | Free and Serverless: AWS and Google Cloud. Dedicated: AWS, Google Cloud and Azure. You can also subscribe through the AWS, GCP or Azure Marketplace |
Compliance | SOC 2 Type II and ISO/IEC 27001 compliant, GDPR ready |
Support | Free plan: community support through a Slack invite link |
Billing terms | Pay-as-you-go; committing to an annual plan can earn additional credits based on your usage |
Our take
Zilliz Cloud is a good choice if you want Milvus-based vector search as a managed service and plan to grow from a prototype into production controls with one vendor. If your team already runs Milvus itself and has the engineers to tune and maintain it, weigh that effort against the managed price before you switch.
Evaluate Zilliz
Comparisons
Where it stands out
Upgrade path inside one service. The plan comparison lists cross-cluster migration from free and serverless clusters and zero downtime migration, so a prototype can move to a Dedicated cluster without changing vendors.
Broad search features. Hybrid search across multiple vector fields (multimodal, sparse-dense and dense-text), plus filtered, range, grouping and full text search, and tunable consistency levels. Rexera's Sasidhar Janaki says hybrid search alone delivered a 40% accuracy improvement for its AI agent architecture.
Less manual tuning. Zilliz says AutoIndex and its Cardinal search engine pick search strategies with zero manual tuning. Shulex's CTO reports lower operational costs and faster search after moving from open-source Milvus to Zilliz Cloud.
Search over data in your own bucket. An optional access mode works directly on your S3 data (Iceberg, Lance, Vortex, Parquet): the data stays in your bucket while Zilliz builds and serves the indexes, with no copies and no ETL.
Product screenshot
Trade-offs to weigh
Capacity figures assume lean data. Per-CU capacity estimates use vectors with primary keys only. Zilliz says additional scalar fields like IDs or labels may reduce capacity and recommends your own tests.
Estimates need a proof of concept. Zilliz says calculator prices are estimates and may differ from actual costs. It suggests a proof of concept with its $200 free credits to check cost and performance.
The newest features are early. Zilliz Vector Lakebase is in Public Preview and On-demand Compute is marked New, so check their maturity before you build core workloads on them.
Savings claims come from Zilliz. Zilliz claims up to 70% lower total cost of ownership than running Milvus yourself and potential savings of up to 80% compared to other vector databases. These are vendor figures; test them on your own workload.
Pricing and plan conditions
Zilliz's cost calculator estimates a bill from query compute units (CUs) and storage in GB. A CU is a group of hardware resources for serving indexes and search requests. The Free plan includes an allowance of virtual compute units (vCUs), which measure resources used by reads (search, query) and writes (insert, upsert, delete); read and write costs vary by vCU usage.
Zilliz gives one worked example for On-demand Compute: 1 billion 768-dimensional vectors, top-k = 100k, cluster size 64 CU. It lists $9.9 cost per 1K searches (5% cold-start, 95% warm), $0 write cost, and $53.7 storage cost per month for the 1B vectors plus index (2.1 TB).
Scroll sideways to see all columns. With a keyboard, focus this table and use the arrow keys.
| Plan | Published price | What it includes |
|---|---|---|
Free | $0 | Learning and personal projects: 5 GB storage, 2.5M vCUs per month included, up to 5 collections |
Standard | From $0/month (Serverless); from $126/month (Dedicated) | Best for prototypes and testing environments: fully managed vector databases with core APIs; backup, restore and basic monitoring; built-in encryption for data in transit and at rest |
Enterprise | From $197/month (Dedicated) | Best for production applications: 99.95% uptime SLA (the plan comparison lists 99.99% if multi-replica is enabled); audit logs, SSO (SAML 2.0 based), granular RBAC; multi-replica and elastic scaling; private endpoint and VPC peering |
Business Critical | Let's Talk | Best for healthcare, finance and other highly regulated, mission-critical systems: global cluster with high-level availability and disaster recovery; CMEK and full-path in-transit encryption; HIPAA-eligible with enhanced data privacy features and rapid incident response |
On-demand Compute | Pay only for active job runtime, with no always-on compute | Lake-scale search and indexing jobs on zero-copy external data: on-demand query jobs and system-managed index builds |
Scroll sideways to see all columns. With a keyboard, focus this table and use the arrow keys.
| Cluster type | Per CU capacity | Search QPS | Search latency | Starting price |
|---|---|---|---|---|
Performance-optimized | 2 million vectors | 500-1500 | 10 ms | From $63 per million vectors / month |
Capacity-optimized | 8 million vectors | 100-300 | 50-100 ms | From $16 per million vectors / month |
Tiered-storage | 40 million vectors | 10-50 | 100-1000ms | From $5 per million vectors / month |
Public signals
About these signals
Verified factual signals from public sources. They indicate observable activity or interest, not total adoption, product quality, or cost.
909.8k PyPI weekly downloads222.1k npm weekly downloads0 vulnerabilities across 2 packages
See all signals from 5 sources
npm · @zilliz/milvus2-sdk-node@3.0.6 · PyPI · pymilvus@3.0.2
Alternatives
Milvus (open source). The same open-source vector database, which you run in your own environment. Zilliz says running Milvus yourself requires significant investment in skilled engineers, performance tuning and ongoing maintenance. Choose it when you have that team and want to operate the database yourself.
Alternatives to Zilliz
The reviewed substitutes for Zilliz among the vector databases, and what would make each one the better answer.
Direct alternatives
Reviewed substitutes: products bought for the same job, where a team picks one.
- Milvus
- Milvus and Zilliz offer the same vector database engine for embedding search: Milvus is the open-source project, and Zilliz sells it as a fully managed cloud service. A team that wants to run and tune its own clusters picks Milvus, while a team that prefers the creators of Milvus to handle deployment, scaling and upkeep picks Zilliz.Applies to: Choosing between two products of the same kind for one job.
- Pinecone
- Pinecone and Zilliz are fully managed vector database services, so teams get similarity search as a cloud product. Pinecone runs its own proprietary engine, while Zilliz runs Milvus, the open-source database built by the same company, which keeps a path open to self-hosting.Applies to: Choosing a vector store for embedding search in a retrieval or agent application.
- Weaviate
- Weaviate and Zilliz offer vector databases that teams can buy as managed cloud services for embedding search. Weaviate is its own open-source database with hybrid keyword and vector search, while Zilliz is the fully managed service for Milvus, run by the company that created Milvus.Applies to: Choosing between these two for the vector databases decision.
- pgvector
- pgvector and Zilliz both store embeddings and answer similarity queries for AI applications. pgvector is an open-source PostgreSQL extension that teams install in their own database or get from a managed Postgres service, while Zilliz is a fully managed cloud service for the Milvus vector database, built for very large collections.Applies to: Choosing between these two for the vector databases decision.
- LanceDB
- Zilliz and LanceDB give AI teams a vector store for search, RAG and agents. Zilliz is a fully managed Milvus service that takes deployment and scaling off the team, while LanceDB is open source and keeps multimodal, versioned data in the Lance format on disk or object storage.Applies to: Choosing between these two for the vector databases decision.
Other approaches
A different approach to the same problem. Each substitutes only for the workload named beside it.
- Qdrant
- Both are managed vector search services; Zilliz is hosted Milvus, Qdrant is Rust-native with a free tier.Applies to: Managed vector similarity search for retrieval in an AI application.