Milvus
Billion-scale open-source vector DB behind Zilliz Cloud
Overview
Milvus is a CNCF-graduated, open-source vector database built for billion-scale similarity search, with GPU-accelerated indexing and standalone or distributed deployment modes. Zilliz, the company behind it, runs the managed Zilliz Cloud.
Pricing
Pricing shown for reference only. These figures reflect RECATOOLS research as of 11 Jul 2026 and may be out of date or incomplete. This is not financial or purchasing advice — always confirm the current price on the provider’s official website before making any decision.
- 5GB storage
- ~100 CU-hours per month
- Auto-scaling compute units
- Storage at $0.04/GB/month (from Jan 2026)
- No idle compute cost
- Continuously running CUs
- Predictable monthly cost
- Higher uptime guarantees
- Custom contracts
- Dedicated support
Use cases
What you can produce with Milvus
- Billion-scale ANN search with GPU-accelerated indexing
- Standalone and distributed deployment modes
- CNCF-graduated governance
- Multiple index types (HNSW, IVF, DiskANN) with tunable recall/latency trade-offs
- Self-hosted (Apache 2.0) or managed Zilliz Cloud
- Milvus Lite for local/embedded development
ASEAN Perspective
Milvus in Southeast Asia
ASEAN-region availability and pricing notes coming soon. Drop the editorial team a note via /contact/ if you can supply local context (Singapore/Malaysia/Indonesia/Thailand/Vietnam).
Milvus is the vector database teams reach for once scale actually matters — GPU-accelerated indexes, distributed deployment, and a CNCF-graduated governance model back a claim of 10,000+ enterprise users and 100 million+ downloads. G2 reviewers in fintech, legal AI and document Q&A describe it as the lowest-friction path from prototype to production once volumes climb past what Chroma or a single Qdrant node handle comfortably. The trade-off is a steep learning curve: index selection and parameter tuning (HNSW vs. IVF vs. DiskANN, and where to trade recall for latency) genuinely require trial and error, and reviewers ask for better built-in guidance here. Zilliz Cloud, the managed option, moved to compute-unit billing in 2025-2026 with storage standardized at $0.04/GB/month across providers from January 2026 — usable, but you need to model usage rather than pick a flat plan. GitHub stars crossed 40,000 in 2026, among the fastest growth in the category. Built for engineering-heavy teams; overkill if you're not past a few million vectors.
What people say
Milvus reviewers on G2 skew toward teams running vector search at genuine scale — fintech, legal AI, document Q&A — and the recurring theme is that it's the database that doesn't fall over once you're past the prototype stage that trips up lighter options. Reviewers specifically cite vector similarity plus metadata filtering holding up on datasets with millions to billions of vectors, and GPU-accelerated indexing is called out as a differentiator versus CPU-only competitors when query latency matters.
The project crossed 35,000 GitHub stars in mid-2025 and passed 40,000 by 2026 — milvus-io/milvus now sits at roughly 45.2k stars, ahead of every other open-source vector database by a wide margin — and Zilliz (the company behind it) says more than 10,000 enterprise teams run it in production, with 100+ million cumulative downloads. As a CNCF-graduated project, governance is more formalized than most vector-DB competitors, which matters to enterprise buyers doing vendor risk review.
The consistent complaint across reviews is the learning curve. Choosing between HNSW, IVF and DiskANN indexes, tuning parameters for the accuracy/latency trade-off, and understanding when distributed mode is actually necessary versus standalone — reviewers describe this as requiring real trial and error rather than being obvious from the docs. A few reviews specifically ask for better built-in tooling around index selection and visibility into what the index is doing.
On pricing, Zilliz Cloud (the managed Milvus offering) restructured around Compute Units in 2025, roughly one vCPU plus 4GB RAM per CU, billed at $0.096/CU-hour on Serverless with storage standardized at $0.04/GB/month across AWS, Azure and GCP starting January 2026. Dedicated tiers start near $99/month; a free tier gives 5GB storage and roughly 100 CU-hours a month, enough for real evaluation. Reviewers running cost comparisons say Serverless works out cheaper for spiky workloads since idle compute isn't billed, while Dedicated is more predictable for steady traffic.
No funding activity beyond the disclosed $113M across Zilliz's rounds (last public raise: a $60M Series B extension in 2022) shows up in the record for 2025-2026.
Summary of public user & expert reviews, compiled by RECATOOLS.
About this listing
This entry was compiled from publicly available data including Milvus's official website, press releases, documentation, and reputable third-party publications. RECATOOLS is not affiliated with Milvus unless explicitly stated.
Third-party AI tools update their pricing, features, availability, and policies frequently. Information here may be outdated by the time you read this — we make reasonable efforts to keep listings current, but cannot guarantee absolute accuracy.
For the latest details, please refer to Milvus directly →
Spotted something out of date? Suggest an update →
Milvus in the news
More in Code & Dev Tools
