Affiliate disclosure: ToolBistro may earn a commission from some links, at no extra cost to you. Facts come from official sources; we do not publish fabricated testing or ratings.

API Pricing

What does the Brave Search API cost in 2026?

Brave Search API charges $5 per 1,000 requests on prepaid credits as of November 2026, with $5 in free credits every month and no subscription. The Answers endpoint costs $4 per 1,000 requests plus $5 per million input and output tokens. Storage rights and custom capacity sit behind an Enterprise plan.

Key facts

What matters

  • The Search endpoint is $5 per 1,000 requests at 50 queries per second; the $5 monthly credit covers roughly 1,000 requests.
  • The Answers endpoint is $4 per 1,000 requests plus $5 per million input and output tokens, capped at 2 queries per second.
  • Brave's FAQ says storing results or training models requires a custom plan, so the self-serve price is not the price for a training pipeline.
  • Google's Custom Search JSON API is closed to new customers and existing users must move by January 1, 2027; it charged the same $5 per 1,000 queries.

What the Brave Search API costs per endpoint

Brave sells one API key across four products, and the price depends on which endpoint you call. Per the plans section of the official pricing page, checked on November 17, 2026:

  • Search, $5 per 1,000 requests. Returns complete results (URLs, text, news, images) with LLM context, Goggles reranking, extra alternate snippets and schema-enriched results. Capacity is 50 queries per second.
  • Answers, $4 per 1,000 requests plus $5 per million input/output tokens. Grounded answers with citations, streaming and OpenAI SDK compatibility. Capacity is 2 queries per second.
  • Enterprise, custom. Full-funnel zero data retention, custom agreements and NDAs, invoicing, and search with storage rights.

The billing model is prepaid credits rather than a monthly seat. Brave's own FAQ explains why the signup form asks for a card: "Brave uses prepaid credits, not subscriptions, so you're only charged for credits you choose to buy." Every plan card also states that $5 in credits is applied automatically each month.

A worked example at 500,000 searches a month, using only the published rates: Brave bills $2,500 at $5 per 1,000, minus the $5 monthly credit, so $2,495. Tavily's pay-as-you-go rate of $0.008 per credit would come to $4,000 if each search consumed one credit. Exa Lite would be $2,000. Google's Custom Search JSON API is not an option at that volume at all, because its 10,000 queries per day ceiling caps a month at roughly 300,000 queries no matter what you are willing to spend.

What $5 in monthly credits actually buys

At $5 per 1,000 requests, the recurring credit is worth about 1,000 Search calls a month, which is a prototyping budget rather than a production one. A chatbot that averages 10 searches per conversation burns through it in 100 conversations.

The comparison that matters for anyone leaving Google is the free allowance. Google's Custom Search JSON API gave 100 queries per day, roughly 3,000 a month, before charging $5 per 1,000 up to 10,000 queries per day. Brave's recurring credit is a third of that on request count, so a side project that lived inside Google's daily quota will start paying on Brave sooner. Brave offsets part of that with a 50 queries per second ceiling instead of a 10,000 queries per day cap, which matters if your traffic arrives in bursts instead of spread across the day.

Brave Search API vs Google, Tavily and Exa on price

The headline number is no longer where these APIs differ: Brave and Google both landed on $5 per 1,000 requests. The differences are in free allowance, rate limits, what a request includes, and whether you are allowed to keep the results.

Tavily prices in credits: pay-as-you-go at $0.008 per credit, or $30 per month for 4,000 credits, with 1,000 free credits each month and no card required. Tavily's FAQ notes that credit consumption varies by request type, so extract and crawl calls cost more than a plain search. Exa prices per endpoint: $4 per 1,000 Lite searches, $7 for Auto, $12 for Deep, plus $1 per 1,000 pages for contents and $2 per 1,000 requests for monitors, with $10 of credits monthly on the free Starter tier.

Read the units before comparing. Brave quotes per request with no token charge on Search, while Answers adds $5 per million tokens on top of the $4 per 1,000. A vendor quoting $4 looks cheaper than Brave's $5 until you add the model bill.

The catch: storage rights and the 2 QPS ceiling

Two details decide whether the self-serve price is usable for your workload.

Storage rights are not included. Brave's FAQ answers "Can I store results or use them to train models?" with "Yes, with a custom plan" and points to sales for search with storage rights. If you cache results for evaluation sets, fine-tuning, or a persistent index, the $5 per 1,000 rate does not cover your use case, and you should budget for a negotiated contract instead. This is the single most expensive thing to discover after building on the cheap tier.

Answers is throttled at 2 queries per second. That is 120 queries per minute, fine for a chat product and useless for a backfill job that needs to answer 50,000 pages overnight. Bulk grounding work belongs on the Search endpoint at 50 QPS, with your own model downstream.

Brave's compliance claims are also worth reading closely. The page states SOC 2 Type II attestation and an architectural zero data retention design, meaning queries terminate on infrastructure Brave operates rather than on a third party's. Enterprise adds full-funnel ZDR, custom NDAs and invoicing.

Migrating off Google Custom Search before January 1, 2027

Google's own documentation now opens with a notice: the Custom Search JSON API is closed to new customers, and existing customers "have until January 1, 2027 to transition to an alternative solution." Google points small-domain use cases at Vertex AI Search and asks full web search users to contact sales.

For a team already on the JSON API, the swap is mostly mechanical: replace the endpoint, move the key into an X-Subscription-Token header, and remap the response fields (Brave returns its own result objects rather than Google's items array). The budget line barely moves at $5 per 1,000 on both sides, so the migration cost is engineering time and re-testing relevance, not a price increase.

Two things to check before you commit. First, whether your code path stores results, because that pushes you into a custom Brave plan rather than the self-serve card. Second, whether your query volume fits inside 10,000 per day today, since Google's cap was a hard ceiling that Brave does not impose at 50 QPS.

Should you read Brave's own benchmark chart?

Brave publishes an accuracy-versus-cost chart from FreshQA claiming 85.3% accuracy at $18 per 1,000 questions, against $49 for Parallel Advanced, $84 for Exa and $89 for Tavily. The page labels the method as "Brave's own published runs." Treat that as a vendor claim with a published methodology, not an independent result: the all-in figure includes model tokens, so it measures a full grounding stack rather than the search API alone, and no competitor has signed off on the arithmetic.

The latency list on the same page credits AIMultiple: Brave 669 ms, Tavily 998 ms, Exa 1,200 ms, Firecrawl 1,335 ms, SerpApi 2,400 ms, Perplexity 11,000 ms. Useful as a directional signal from a third party, not as a guarantee for your query mix. Run 200 of your own real queries against two vendors before you sign anything; that sample will tell you more than either chart.

Who should pick Brave, and who should not

  • Pick Brave if you are replacing Google's Custom Search JSON API, want one key across web, news, image and video search, and can live inside a prepaid credit balance.
  • Pick Brave Answers if you would rather buy grounded answers with citations than wire up a model plus a search call yourself, and your throughput fits 2 queries per second.
  • Pick Tavily if your pipeline leans on extract and crawl as much as search, since it sells those endpoints under one credit system.
  • Pick Exa if you need semantic or neural retrieval rather than keyword results and can absorb $7 to $12 per 1,000 for Auto or Deep search.
  • Skip all of them for training data. Cheap self-serve search tiers generally restrict storing results, and Brave states storage rights require a custom plan.

At a glance

Provider / endpointPrice per 1,000 requestsFree allowanceNotes
Brave Search$5$5 credits monthly (about 1,000 requests)50 queries per second; includes LLM context
Brave Answers$4 + $5 per 1M tokens$5 credits monthly2 queries per second; grounded answers with citations
Google Custom Search JSON$5100 queries per dayClosed to new customers; migrate by Jan 1, 2027
Tavily$0.008 per credit1,000 credits monthly$30/mo plan includes 4,000 credits; credits vary by endpoint
Exa$4 Lite / $7 Auto / $12 Deep$10 credits monthly on StarterContents $1 per 1,000 pages; monitors $2 per 1,000 requests

FAQ

Is the Brave Search API free?

There is no free tier in the classic sense. Brave applies $5 in credits every month, which at $5 per 1,000 requests covers about 1,000 Search calls, and bills the rest from prepaid credits with no subscription.

Is Brave Search API cheaper than Google Custom Search?

No, the paid rate is the same: $5 per 1,000 requests at both. Google gave 100 free queries per day (about 3,000 a month) against Brave's credit of roughly 1,000, but Google closed the JSON API to new customers and requires existing users to migrate by January 1, 2027.

Can I store Brave Search API results or train a model on them?

Only with a custom plan. Brave's FAQ says storage and training use cases require a custom plan arranged through sales, so caching results indefinitely is outside the self-serve $5 per 1,000 terms.

Related reading

web search API pricing compared for AI agents, SerpApi alternatives

Sources