Serper vs Tavily (2026): Pricing, Output and Which to Use

Serper is far cheaper per search: $1.00 per 1,000 on its smallest pack and $0.30 on its largest, against Tavily's $8.00 pay as you go and $5.00 on its $500-a-month Growth plan. Depending on which tiers you compare, Serper costs 5 to 26.7 times less per basic search. They return different things, though. Serper gives you Google's results page as JSON: organic results, knowledge graph, People Also Ask and related searches. Tavily gives you results with snippet chunks, and the same request can add each page's raw content, with no extra credit charge listed, and a generated answer. Serper's page content comes from a separate Webpage endpoint whose price is shown only after sign-in. Pick Serper for cheap Google results. Pick Tavily when an LLM needs to read the pages.

Prices checked on the vendors' own pages: Serper on 19 October 2026 at serper.dev, Tavily on 1 October 2026 in Tavily's API credits docs.

At a glance

Price per 1,000 searches, with the working

Serper has four packs: $50 for 50,000 credits, $375 for 500,000, $1,250 for 2.5 million and $3,750 for 12.5 million. Divide each price by its credits in thousands and you get $1.00, $0.75, $0.50 and $0.30 per 1,000. Tavily has two rates: $8.00 per 1,000 credits pay as you go and $5.00 on the Growth plan.

Each cell below is Tavily's rate divided by Serper's, for a basic Tavily search (1 credit):

So the gap runs from 5× (Tavily Growth against Serper's smallest pack) to 26.7× (Tavily pay as you go against Serper's largest). Like for like, entry against entry is 8× and volume against volume is 16.7×.

Two things move these numbers:

• Tavily's advanced depth costs 2 credits, which doubles every ratio: $16.00 and $10.00 per 1,000, so 10× to 53.3×.
• Serper's 6-month expiry. The $0.30 rate only holds if you use 12.5 million credits within 6 months, about 2.08 million a month. Use less and some of the pack expires, which raises your real rate.

And one thing the table leaves out: what you get per search is not the same. If you need page text, Tavily can include it in the search at no extra credit charge listed. With Serper you pay for search plus the Webpage endpoint, and since that endpoint's price isn't public, we can't compute the combined cost.

What 10,000 and 100,000 searches a month cost

Assuming basic-depth Tavily searches, a steady monthly volume, and no free credits counted:

At 100,000 a month the Growth plan's $500 buys exactly the 100,000 credits you need ($500 ÷ $5.00 × 1,000). With advanced depth that doubles to 200,000 credits, which is $1,600 pay as you go.

On Serper, the pack to buy at both volumes is one you'll finish inside 6 months. A bigger pack has a lower headline rate, but most of it would expire.

Free tiers

Serper gives 2,500 free queries once, with no card. Tavily's free Researcher plan gives 1,000 credits every month, also with no card.

Serper's is bigger on day one. Tavily's passes it after three months (3,000 against 2,500) and keeps going. Tavily's credits buy 1,000 basic searches a month, or 500 at advanced depth. For a prototype you'll test hard for a week, Serper's allowance goes further. For a side project that runs a few hundred searches a month, Tavily's free plan covers it every month.

Output: what one request returns

Serper returns Google's results page as JSON: organic results, knowledge graph, People Also Ask and related searches. Beyond web search it covers Images, News, Maps, Places, Videos, Shopping, Scholar, Patents and Autocomplete, with country and language targeting. Serper's own FAQ says "All queries are returned in 1 to 2 seconds"; that's their claim, not our measurement.

Tavily returns results with snippet chunks. include_raw_content adds the raw content of the result pages and include_answer adds a generated answer. Its RAG helpers, answer synthesis and follow-up questions, are built for apps where an LLM reads the output.

If your code needs to know what Google shows for a query, including its SERP features, that's Serper. If an LLM is going to read the output and answer from it, Tavily hands it more to work with in one call.

Content handling

This is where the per-search comparison stops being the whole story.

• Tavily: set include_raw_content on the search and the page content comes back in the same response, with no extra credit charge listed. For URLs you already have, the separate /extract endpoint costs 1 credit per 5 URLs at basic depth (roughly $1.60 per 1,000 pages pay as you go) and 2 credits per 5 at advanced.
• Serper: the search returns the results page only. Page content is a separate Webpage endpoint, and its credit price is shown only after sign-in. Budget for separate Webpage calls, at a rate you'll need to check in your account.

Integrations

Both work with LangChain. Serper has docs pages for LangChain (GoogleSerperAPIWrapper), CrewAI (SerperDevTool) and Haystack. Tavily has mature LangChain and LlamaIndex integrations. If you're on CrewAI or Haystack, Serper's docs cover you directly. If you're on LlamaIndex, Tavily's integration is the established one.

Which should you pick

Pick Serper if:

• You want Google results as JSON for far less than Tavily charges: $0.30 to $1.00 per 1,000.
• You need SERP features (knowledge graph, People Also Ask, related searches) or other Google search types such as News, Maps, Shopping or Scholar.
• You need country and language targeting.
• You already fetch and parse pages yourself, or only need snippets.

Pick Tavily if:

• An LLM or RAG pipeline reads the results and you want page content and an answer from the same request.
• You want RAG helpers such as answer synthesis and follow-up questions.
• You use LlamaIndex, or want Tavily's mature LangChain integration.
• Your volume is small and steady enough for 1,000 free credits a month.

Where each is weaker: Tavily costs 5 to 26.7 times more per basic search, and advanced depth doubles that. Serper leaves page content to a second endpoint with no public price, and its credits expire after 6 months.

A third option: Serpex

If you want search results and page content in one call without paying Tavily's per-search rate, Serpex is a third option. One request returns results as JSON, and with include_content: true it also returns the page content of the top 5 or 10 results as Markdown, at a published price: a search costs $0.50 to $0.80 per 1,000 depending on top-up size, and content costs at most 3 credits for 5 pages or 6 for 10, billed only on pages delivered. That's $3.00 to $4.80 per 1,000 requests with 10 pages of content. Content is best-effort: it arrives for roughly 70% of requested results, and pages that are blocked or disallowed by robots.txt return a content_error instead. Serper's $0.30 pack is lower than any Serpex list rate for plain search, and Serpex doesn't return a generated answer the way Tavily does. New accounts get 200 free credits. More detail: Serper alternative, Tavily alternative, and the wider price table in Search API Pricing in 2026.

FAQ

Is Serper cheaper than Tavily?

Yes, per search. Serper costs $0.30 to $1.00 per 1,000 and Tavily $5.00 to $8.00 for basic searches, so Serper is 5 to 26.7 times cheaper depending on the tiers you compare.

Which free tier is bigger, Serper or Tavily?

Serper gives 2,500 free queries once. Tavily gives 1,000 credits every month. Serper's is bigger at the start; Tavily's adds up to more after three months.

Does Serper return page content?

Not in the search response. Page content comes from a separate Webpage endpoint whose price is shown only after sign-in. Tavily can include raw page content in the search itself.

Do Serper and Tavily work with LangChain?

Yes. Serper has a LangChain docs page for GoogleSerperAPIWrapper, plus pages for CrewAI and Haystack. Tavily has mature LangChain and LlamaIndex integrations.

How were these prices checked?

From each vendor's own page: Serper on 19 October 2026, Tavily on 1 October 2026. Per-1,000 figures are price divided by credits. Prices change often, so check before you commit.