> ## Documentation Index
> Fetch the complete documentation index at: https://docs.truscan.co/llms.txt
> Use this file to discover all available pages before exploring further.

# Search tiers

> Trade latency and cost for recall depth.

`tier` selects how hard Truscan works on a query. It changes how many candidates
are recalled and how many pages are enriched, and therefore latency and price.

| Tier             | Price / 1k requests | Recall | Enrichment |
| ---------------- | ------------------: | -----: | ---------: |
| `instant`        |              \$7.00 |     1× |         1× |
| `deep`           |             \$12.00 |     2× |         2× |
| `deep_reasoning` |             \$15.00 |     3× |       3.5× |

`instant` is the default when `tier` is omitted.

## Choosing one

<AccordionGroup>
  <Accordion title="instant: interactive lookups" icon="bolt">
    The default. One pass across the engines, enough for a factual question
    where the answer is on the first page of results. Use it for autocomplete,
    chat turns, and anything a human is waiting on.
  </Accordion>

  <Accordion title="deep: research and coverage" icon="layer-group">
    Doubles both recall and enrichment. Use it when missing a relevant source is
    worse than waiting: competitive analysis, literature sweeps, or questions
    where sources disagree.
  </Accordion>

  <Accordion title="deep_reasoning: maximum evidence" icon="brain">
    Widest recall (3×) and the heaviest enrichment (3.5×). Use it when an agent
    needs to reason over a broad evidence set and you would rather pay once than
    re-query.
  </Accordion>
</AccordionGroup>

## Aliases

Two aliases normalize to `instant`, so existing integrations keep working:

| You send | Treated as |
| -------- | ---------- |
| `fast`   | `instant`  |
| `auto`   | `instant`  |

The `tier` field in the response is always the canonical name.

## Cost

Tier sets the base rate. Two things add to it:

* results beyond the **10 included** per request, at `$1.00 / 1k results`
* pages enriched when `enrich` is `true`, at `$1.00 / 1k pages`

A `deep` search returning 10 enriched results costs the \$12.00/1k base plus
10 pages of extraction. To price a call before you make it, use
[`POST /api/billing/quote`](/billing/pricing#quoting-before-you-spend).

<Tip>
  Repeating an identical query is served from cache and is **not charged**.
  `tier` is part of the cache key, so the same query at a different tier is a
  new, billable search.
</Tip>