Weaviate is an open-source, AI-native vector database built by Weaviate B.V. It stores and searches unstructured data such as text, images, audio, and video using vector embeddings alongside structured attributes. Teams use it to build semantic search, retrieval-augmented generation, and stateful AI agent applications.
The system converts unstructured objects into dense vector representations, either through built-in model integrations or custom external embeddings. It indexes these vectors with HNSW or Flat indexing, enabling k-nearest neighbor search, hybrid keyword-vector search using BM25, and metadata filtering in one query engine.
Named vectors let a single object store multiple embeddings from different models or modalities. Engram, a managed memory layer for autonomous AI agents, reached general availability. Deployment options span self-hosted open source, a free managed tier, and paid cloud plans with shared or dedicated infrastructure.
Weaviate B.V. is based in Amsterdam and was founded in 2019 by Etienne Dilocker and Bob van Luijt. The company has raised funding across multiple rounds, including a Series C led by Battery Ventures and Zetta Venture Partners, plus a 2026 strategic investment from Ricoh’s corporate venture fund.
Pricing
Weaviate offers a free forever plan capped at 100,000 objects, 1 GB memory, and 10 GB disk. The Flex plan starts at $45 per month on a pay-as-you-go basis, and Premium starts at $400 per month on a prepaid contract with shared or dedicated deployment. Weaviate Embeddings are billed separately starting at $0.025 per 1 million tokens, and the Query Agent costs $30 per month for 4,000 requests after a free tier of 1,000 requests. Self-hosted open source deployment remains free under the BSD-3-Clause license.
* Disclaimer: Please note that pricing information may not be up to date. For the most accurate and current pricing details, refer to the official website.
Key Features
- ✓
Hybrid search merges BM25 keyword matching with dense vector search
- ✓
Named vectors store multiple embeddings per object from different models
- ✓
Engram delivers managed long-term memory for AI agents, now generally available
- ✓
Modular vectorizer integrations connect to OpenAI, Cohere, and Hugging Face
- ✓
Tunable indexing supports HNSW plus compression methods like RQ-8
- ✓
GraphQL, REST, and native SDKs cover Python, TypeScript, Go, and Java
Use Cases
Retrieval-Augmented Generation Pipelines
Teams retrieve relevant proprietary documents before passing them to an LLM. This grounds generated answers in verified data and reduces hallucinated responses.
Enterprise Semantic Search
Weaviate replaces exact-match search across product catalogs and internal knowledge bases. Hybrid search blends keyword precision with semantic relevance for more accurate results.
Long-Term Memory for AI Agents
Engram extracts facts from agent interactions and consolidates them into structured memory. Agents retrieve preferences and past context across sessions.
Multi-Modal Media Discovery
Images, video frames, and text metadata are indexed in a shared vector space. Users search visual assets with natural language or find similar images.
Multilingual Document Retrieval
Cross-lingual embeddings let teams query multi-language libraries from a single language. This helps global platforms surface relevant documentation without translation.
Strengths & Weaknesses
Strengths
Combines BM25 keyword search and vector search in one query engine with adjustable weighting.
Automates vector generation through built-in model integrations, removing manual embedding pipelines.
Supports named vectors, so one object can hold embeddings from multiple models.
Offers a free forever tier alongside pay-as-you-go and prepaid cloud plans.
Is SOC 2 Type II audited, with HIPAA compliance available on Enterprise Cloud (AWS).
Weaknesses
Premium plan pricing starts at $400 per month, a jump from Flex’s $45 minimum.
Self-hosting at scale on Kubernetes requires dedicated infrastructure expertise.
Automatic vectorization depends on third-party model APIs, adding potential latency or rate limits.
Free tier is capped at 100,000 objects and one collection, limiting production use.
Who Is This For?
AI and ML Engineers: building RAG pipelines and semantic search features that need fast vector retrieval and SDK support.
Enterprise Systems Architects: needing dedicated cloud deployment for compliance and data hosting requirements.
AI Agent Developers: building stateful agents that rely on persistent memory through Engram.
SaaS Product Teams: building multi-tenant applications that require data isolation across customers on shared infrastructure.
Frequently Asked Questions
What happens if my data grows past the free plan limits?
The free plan is capped at 100,000 objects, 1 GB memory, and one collection. You need to upgrade to Flex or Premium to scale beyond those limits.
Can I run Weaviate without using Weaviate Cloud?
Yes. Weaviate is open source under the BSD-3-Clause license and can be self-hosted using Docker or Kubernetes.
What is Weaviate Engram used for?
Engram is a managed memory service that helps AI agents extract facts, resolve inconsistencies, and retrieve context across sessions. It reached general availability.
Does Weaviate support multiple vectors per data object?
Yes, through named vectors, a single object can hold embeddings generated by different models or modalities.
What query interfaces does Weaviate support?
Weaviate supports GraphQL, REST endpoints, and client SDKs for Python, TypeScript, Go, and Java.
How is Flex plan pricing calculated?
Flex starts at $45 per month on a pay-as-you-go basis, priced by vector dimensions and storage used.
What separates the Flex and Premium plans?
Flex is monthly pay-as-you-go on shared infrastructure with 99.5% uptime. Premium is a prepaid contract starting at $400 per month with shared or dedicated deployment and up to 99.95% uptime.
Is Weaviate compliant with security standards like SOC 2 or HIPAA?
Weaviate Cloud is SOC 2 Type II audited. HIPAA compliance is available on Enterprise Cloud deployed on AWS.
Which embedding providers does Weaviate integrate with natively?
Weaviate connects to OpenAI, Cohere, Hugging Face, Google Vertex AI, and AWS Bedrock, along with LangChain and LlamaIndex.
Are embeddings and the Query Agent included in the base plan price?
No. Both are billed separately by usage. Embeddings start at $0.025 per 1 million tokens, and the Query Agent costs $30 per month for 4,000 requests after a 1,000-request free tier.
Weaviate integrates natively with OpenAI, Cohere, Hugging Face, Google Vertex AI, and AWS Bedrock for vectorization and generative modules. It also connects with LangChain and LlamaIndex for building RAG applications.