Exa

We sat down with ⁨@3blue1brown to discuss our research vision of "Attention over all the world's information."

Plus why knowledge should not live in model weights, the compression trick behind cheaper retrieval, and why we hire through work trials.

1:20 – What is “attention

Even the top search agents set to the highest effort modes cannot find 1/3 of the verified information.

Read the full research: exa.ai/blog/atlas-ben…

ATLAS tasks require many sources to retrieve fresh or niche information.

Questions are synthetically generated through an automated pipeline based on real-world search trends, covering 20+ topic categories.

ATLAS reflects the value of search far more than existing search benchmarks.

When intentionally holding back search results from reaching the agent, ATLAS questions are less answerable.

Existing evals have become saturated. Leading search agents score 93% on BrowseComp and 95% on DeepSearchQA.

And agents have memorized their answers:

We built ATLAS, a web search benchmark for how agents actually search today.

Existing agentic search benchmarks are often outdated, contaminated, or saturated. ATLAS is designed to separate strong search from weak search.

Learn more: exa.ai/blog/atlas-ben…

Exa Places is live across all our endpoints at no additional cost!

Learn more
exa.ai/blog/places-se…

Exa Places significantly improves the accuracy of personal agents that are performing tasks with local intent.

It has the highest accuracy across search providers we evaluated.

We just added 211 million restaurants, museums, and parks to Exa's index.

Give your AI agents access to search over all the places in the world with structured data including opening hours, reviews, and phone numbers.

You can use Agent Ultra in the Exa API today!

Learn more - exa.ai/blog/exa-agent…

We find that on many tasks Agent Ultra is significantly cheaper than other providers

Agent Ultra uses a custom research harness that efficiently uses subagents, code execution, and Exa's token efficient content highlights.

This makes Agent Ultra state-of-the-art on web research, while being Pareto dominant on cost and quality.

Introducing Agent Ultra - a step change in deep research

Agent Ultra orchestrates swarms of agents to perform exhaustive research, build comprehensive lists, and answer questions requiring thousands of sources.

In both evals and vibes, it's state of the art

Jev with Exa is INSANE.

> Jev without websearch confidently gives wrong outputs
> Jev with websearch is literally much more accurate

Try Jev (with Exa websearch) for free 👇

Snapshots are available to try in Exa today.

Learn more: exa.ai/blog/exa-snaps…

We built a time machine for the web.

Introducing Exa Snapshot: an index of 400 billion historical snapshots of webpages that lets you search as if it's the past.

Snapshot is already being used for backtesting prediction models, RL at labs, exploring the pre-AI web, and more.

Exa is super token efficient! We made a calculator that shows how it compares to OpenAI/Anthropic native search:

Exa Agent can now answer questions with live Polymarket data.

Odds, price history, order books, trader positions, and more.

We’re releasing a research preview of “dynamic highlights”:

Given a query and set of pages, a custom model jointly scans the whole set to filter out redundant and irrelevant info, reducing token usage 95% on average.

Available today: exa.ai/blog/dynamic-h…