Public Beta for LanceDB Cloud is now available! Try it out!https://cloud.lancedb.comlink-arrow

The Database for Multimodal AI

Designed from the ground up for AI to store and query all of your data from embeddings to text, images, audio, and video. LanceDB delivers blazing fast retrieval on billions of vectors from object storage, accelerates training or fine-tuning, and supports distributed pre-processing for petabyte-scale multimodal datasets.
Performance seen in production

Largest dataset in training

20 Petabytes

Largest single table indexed

10 Billion+ rows

Highest QPS on a single table

20000+ queries / second

Max Ingestion

4GB / second

Learn more about Performance - LanceDB link-arrow

The easiest way to manage multimodal data, from experiment to production
Installs in seconds as an embedded database for rapid prototyping. Scales to petabytes in production on object storage for everything from embeddings and metadata to raw image, audio, and video bytes.
Vector search is critical infrastructure that allows us to better serve our users. We evaluated multiple solutions and LanceDB was the only one that could meet the high-traffic and large scale requirements we had. We couldn't be happier with our decision.
Nadia Ali
CFO
The only solution to serve tens of billions vectors without breaking the bank. When we look at the price-performance curve, no one came close to Lance.
Jiaming Song
Chief Scientist
Law firms, professional service providers, and large enterprises rely on Harvey to process a large number of complex documents in a scalable and secure manner. LanceDB has done a great job in helping us meet those demands, and their search/retrieval infrastructure is a key driver for our platform.
Gabriel Pereyra
President & Co-Founder
Migrating full-text search from ElasticSearch to LanceDB reduced our p90 latency by over 90%. The integration with Spark allows us to continue iterating on this system faster than ever!
Noah Shpak
Member of Technical Staff
Hex uses Lance to power semantic search across thousands of users and hundreds of vector spaces with frequent updates throughout the day
Bryan Bischof
Director of Engineering, AI
LanceDB isn’t just a tool—it’s our cheat code. We’re shipping crazier ideas, faster, with minimal operational overhead, and our players love it.Read the case study.
Xiaoyang Yang
VP of AI, Data and Security
Lance has helped us streamline our data analysis process, allowing us to quickly and efficiently identify valuable data to enhance our autonomous driving technology.Read the case study.
Fei Chen
Director of Data Infrastructure
The only embedded vector search solution available for TypeScript, LanceDB's open source solution made it easy for us to build scalable and distributed retrieval for RAG, We are really happy with how the system has scaled.
Joshua Ma
Head of AI
Thanks for all the work that you do! When I found LanceDB it was exactly what we needed, and has played its role perfectly since then :)
Nate Sesti
CTO
LanceDB made it so much easier for us to develop and manage our dozens of different kinds of embeddings. It works out of the box with all of our familiar tools like pandas or Spark, and it stores data directly on S3. LanceDB has enabled us to match a user with exactly the stories they're interested in.
Marios Assiotis
CTO
Of all instances of AnythingLLM, 96% use our default, and recommended, vector database, LanceDB for storage and retrieval of their documents. LanceDB allows all users the best of both performance for vector search as well as privacy - running totally on the user's laptop.Read the case study.
Timothy Carambat
Founder CEO @Mintplex
We looked through a variety of vector search solutions when building Shaped and Lance was the clear winner. We chose it because of how well it scaled in our multi-tenant architecture, the latency benefits for our real-time use-cases and the flexibility of the familiar DataFrame interface.
Tullie Murrell
CEO & Co-Founder
Code is constantly changing, making reproducing previous agent runs incredibly challenging. LanceDB's out-of-the-box versioning and time-travel capabilities has made it significantly easier for us to reproduce, experiment, and evaluate on historic data.
Devin Stein
Founder
Unparalleled Scalability for Multimodal AI
LanceDB is a massively scalable all-in-one solution for AI retrieval, data pre-processing, and accelerated training.
Performance at Scale
icon
Search on billion-vector indices in milliseconds; supports GPU-accelerated indexing
Cost Effective Scalability
icon
Grow your scale 10-100x at a fraction of the cost; full compute-storage separation using object storage with advanced auto-scaling and caching.
Better Retrieval; Better AI
icon
Easily improve the accuracy of your AI retrieval with fast filters, hybrid search, reranking, late-interaction, and other advanced techniques.
Unified Multimodal Storage
icon
Store all of your AI data from embeddings and metadata to the actual audio, image, and video bytes. Multiple systems in one. Fights S3 rate limits for you.
Training and Pre-processing
icon
Distributed pre-processing and accelerated training cache. Native support for LLM-as-UDF in your data pipeline.
Powered by Lance Format
icon
New open source columnar standard for multimodal data. 100x faster than parquet. Better schema evolution than Iceberg. Trusted by cutting-edge multimodal AI labs.
Trusted by enterprises
Deployed at production-scale with the most challenging requirements across multimodal generative AI, code-gen, legal, healthcare, fintech, robotics, streaming, e-commerce and much more
LanceDB Cloud is SOC2 Type II and HIPAA certified.