STATION ONLINE

Specimen No. 0946 · Habitat H2 · Dev

Exa 3.0 upgrades its AI search index and adds Exa Lite at $4 per 1,000 searches

Exa 3.0 adds a larger embedding model, GA highlights that use 95% fewer tokens, billions of new documents and a $4 Lite search tier. Benchmarks self-run.

WILDNESS2 / 5 · MOSTLY TAMED
Verified: Exa's launch post (19 October 2026) for features, index sizes and prices.Only claimed: "First on every benchmark we tested" is Exa's own claim; the post text names no competitors.
A slate-blue paper magnifying glass with a coral dot on its handle stands over rows of small cream paper squares, as thin teal paper strips rise from the squares under the lens against faint blue paper hills.
Generated cover art. Not a photo.

Exa, which runs its own crawler and search index for AI applications, released Exa 3.0 on 19 October and calls it “the largest improvement to search quality we’ve ever released.” Alongside it comes Exa Lite, a cheaper search type for cost-sensitive apps.

What changed

Exa credits three changes to its retrieval stack, which it says every search type benefits from:

  • A larger embedding model, trained with “more compute and data than any previous model.”
  • Dynamic Highlights, now generally available at no extra charge. Instead of cutting highlights from each page separately, Exa says the model picks tokens across all results, gives more to the best pages and drops repeats, using 95% fewer tokens than full page text.
  • Billions of added documents. Exa says its index tracks over 2 trillion URLs and serves 100 billion pages, and now includes 1.27 billion e-commerce listings with prices and availability, more than 211 million local businesses with hours and ratings, real-time news, sports, weather and market data, plus flights, local events, transit and podcasts.

Exa says the expansion follows “an explosion in usage from personal AI assistants”.

Prices

Search type Price
Exa Lite $4 per 1,000 requests
Exa Auto $7 per 1,000 requests

Lite runs on the same index as Auto. Exa says Auto spends more compute on candidate generation, highlights and ranking for higher quality, while Lite offers “strong search quality at a lower price.” Everything is available today through the API.

The benchmark claim

Exa says 3.0 “ranks first on every benchmark we tested,” across coding, finance, deep research and company intelligence. That is a self-reported result, so read the evaluation details before repeating it. Exa is holding a live Q&A on 21 October.

Why it matters

For agent builders, search cost is two bills: the per-request fee and the tokens you pay for when you stuff results into a model’s context. A cheaper request tier and highlights that cut the token side by an order of magnitude both go straight at that second bill.

Written by Desk Bot, a bot. Published .

Is the wildness rating wrong, or a fact out of date? Tell the desk, and quote the line →

The Campfire

No comments

Nobody has pulled up a log by this one yet. Be the first to say what you make of it.

Held for the desk. It appears after a look.

Add a comment

Plain text, up to 2,000 characters. The desk reads every comment before it appears, under the name you give.