Get any slice of the web
Up toBillions
of urls
Up toTrillions
of tokens
Example datasets
Why use Exa
High Quality data
Instead of relying on low quality datadumps like commoncrawl, Exa's crawling infrastructure is optimized to find data from high quality sources.
Powerful Filters
Exa can filter the web based on natural language. You can filter by topic, idea, entity type, length -- pretty much anything you want.
Comprehensiveness
We can customize our crawling to fit any size that fits your needs: from a few million tokens to trillions.
Find financial articles that mention S&P500 companies in thelast year
Use case
Training a large context window model
Challenge
A leading AI company is trying to fine-tune its own LLM to handle long context windows for technical topics like programming, but cannot find high quality, long-form, technical content.
Solution
Using a few natural language queries, Exa is able to find high quality, long-form, technical content in math, science and programming that exactly match the fine-tuning goal of the company.
Impact
>10Btokens delivered
The Exa dataset was delivered quickly, saving the company from dedicating a team of engineers for months to build out a crawling system and filtering algorithm from scratch.
Exa gives real web data and control to your dataset
Features
Types of web data
High-quality sources
Random webpages
Ability to filter data
Customized crawling
Update interval
Every day
Every month
Customer support
Trusted by thousands of developers and companies
“Exa feeds our deep research AI, which helps sales people research their prospects. Without Exa's speed and quality over the web, this would be hard to pull off!”
“Models are only as good as the data they're trained on, and Exa's search allowed us to get high quality data we couldn't find any other way”
“Exa is good, really good. We went from multiple API calls and scraping into a single <1s fast call. The results are way different than traditional search, and way better. Our users love it!”
