What one API request really costs: price per call vs price per 1,000 records

One API request costs whatever the seller charges per call, but that number alone doesn't tell you what the data costs. Divide the price of the call by the records it returned and you get dollars per 1,000 records, which is the figure you can compare across providers. On the six real calls below, ranking by price per call and ranking by price per 1,000 records give different orders.

Six real calls, priced two ways

We made each call through the FetchAPI gateway on 7 October 2026 and counted the records in the answer. The price is what the gateway metered for that call: the provider's per-call price on its entry plan.

Two things change when you switch columns:

• Instagram vs TikTok. Per call, FlashAPI's Instagram call ($0.00099) was half the price of tiktok-api6's call ($0.001998). Per 1,000 records it was about 24% more expensive ($0.0825 vs $0.0666), because it returned 12 posts and the TikTok call returned 30.
• Reddit vs X. Per call, the Reddit search cost 1.7x the X search. Per 1,000 records the gap shrinks to 1.3x, because the Reddit call returned 25 posts to X's 19.

The math

price per 1,000 records = price per call / records per call x 1,000

For the X search: $0.00009 / 19 x 1,000 = $0.0047. To collect 1,000,000 posts you need about 52,632 calls (1,000,000 / 19), so about $4.74 at this price.

Three things make that figure an estimate rather than a quote:

1. Page size varies. We asked tiktok-scraper7 for 20 videos and got 19. A search returns fewer results when fewer match, and later pages are often shorter. Say "about" and measure your own queries.
2. Records aren't all the same. A Google result page from DataForSEO came back with 11 items: 8 organic results, an AI overview, a video block and related searches. Whether that is 8 records or 11 depends on what you need.
3. Pagination costs calls. If an endpoint returns 12 posts a page, 1,000 posts take 84 calls, and the cursor calls are billed like the first one.

Per-call vs per-record pricing models

Sellers don't all price the same way, so convert everything to dollars per 1,000 records before comparing:

With per-record pricing, page size doesn't matter. With per-call pricing it decides everything. AnyAPI uses both: most social endpoints are flat per request, and some are "linear", a base fee plus a fee per result (its instagram.hashtag_analytics is $0.0011 per call plus $0.00187 per result, for example).

The official X price is in the table because it is the ceiling, not because it is comparable: it includes a licence and terms that the unofficial routes don't. We compare the two in Why the same TikTok data costs 100x more from one API than another.

Why the search result is the expensive one

The DataForSEO call stands out: $0.25 per 1,000 organic results, against $0.001 per 1,000 TikTok videos. One Google results page is a fresh query to a search engine, and DataForSEO bills one page of 10 results as one unit ($0.002 in live mode). It also took 12.3 seconds against 1.5 to 3.7 seconds for the social calls. Why search data costs more is its own post: Why search and SEO data costs more than social media data.

What this means for an agent

An agent pays per tool call, so the same arithmetic decides what a task costs. Three habits keep it low:

• Ask for the largest page the endpoint allows. Many list endpoints take a count or limit parameter; the call costs the same whether it returns 10 records or 30.
• Prefer a list call over many lookups. Thirty single-video lookups cost thirty calls; one user-videos call returned 30 videos for the price of one.
• Stop paginating when you have enough. Every cursor page is a full-price call.

Caveats on these prices

• The RapidAPI prices (twitter-api45, tiktok-scraper7, tiktok-api6, reddit-com, FlashAPI) are each plan's monthly price divided by its quota. If you buy the plan yourself and use less, each call costs more. See Subscription quota vs pay per call.
• Through FetchAPI you pay that full-quota price per call without the subscription, times our markup, which is currently 1.0. Failed calls are free.
• Record counts come from one call each. Measure your own query before you budget a large job.

Try it

Run the X search yourself; meta.cost_usd in the answer is the metered price, and the posts are in data:

curl -s "https://fetchapi.co/api/gw/rapidapi-twitter-api45/search.php?query=web%20scraping%20api&search_type=Top" \
  -H "x-api-key: $FETCHAPI_KEY"

Count the records, divide, and you have your own price per 1,000. Every endpoint's per-call price is in the catalog, with platform pages for X, TikTok, Instagram and Reddit, and the same endpoints work as tools over MCP. Get a free key to start.