Tavily vs Exa in 2026

Both calls, one key

You do not have to choose an API or rewrite one. Tavily's call and Exa's call both run on SpecStream, each on its own path and in its own JSON. The address changes, and the key.

# Tavily: before
curl -X POST https://api.tavily.com/search \
  -H "Authorization: Bearer $TAVILY_API_KEY" -H "Content-Type: application/json" \
  -d '{"query": "Who founded Mistral AI and when?", "max_results": 5}'

# Tavily: after
curl -X POST https://api.specstream.ai/tavily/search \
  -H "Authorization: Bearer $SPECSTREAM_API_KEY" -H "Content-Type: application/json" \
  -d '{"query": "Who founded Mistral AI and when?", "max_results": 5}'

# Exa: before
curl -X POST https://api.exa.ai/search \
  -H "x-api-key: $EXA_API_KEY" -H "Content-Type: application/json" \
  -d '{"query": "open source vector databases", "numResults": 10}'

# Exa: after
curl -X POST https://api.specstream.ai/exa/search \
  -H "x-api-key: $SPECSTREAM_API_KEY" -H "Content-Type: application/json" \
  -d '{"query": "open source vector databases", "numResults": 10}'

In the official SDKs it is one argument: TavilyClient(api_key=SPECSTREAM_API_KEY, api_base_url="https://api.specstream.ai/tavily") and Exa(api_key=SPECSTREAM_API_KEY, base_url="https://api.specstream.ai/exa").

What stays the same

The Tavily route reads query, max_results (0 to 20), include_domains, exclude_domains, time_range, topic news with days, start_date, end_date, include_answer and include_raw_content, and returns results[] with title, url, content, score and raw_content. The Tavily migration guide lists every parameter and field.

The Exa route reads query, numResults (1 to 100), includeDomains, excludeDomains, includeText, excludeText, startPublishedDate, endPublishedDate, contents.text and contents.highlights, and returns results[] with id, url, title, publishedDate and text, plus costDollars.total, the price of the call. The Exa migration guide lists every parameter and field.

One price for every call

On SpecStream a search of up to 10 results costs $0.005 whether it is Tavily basic or advanced, Exa auto or deep, with page text or without. Each started block of 10 more results adds $0.006. 100,000 searches cost $500, from credit that never expires and with no monthly plan to renew. A busy call answers 429 with Retry-After and is not charged.

The most accurate web search API in the world, in about one second

The SpecStream Search API scores a 97.9 F-score on all 1,000 official SimpleQA Verified questions, with 978 correct. On WideSearch it scores 64.0 item F1 and 43.4 row F1 on all 200 official tasks, and the Extract API scores 0.908 F1 on WCXB. The scores and the method are at specstream.ai/benchmarks.

A call of up to 10 results without page text runs the fast search. On November 9, 2026 production answered 718 of 720 fast searches: the median was 0.82 s at 1 search a second and 1.07 s at 20 a second, and the 95th percentile stayed between 1.09 and 1.73 s. There is no limit per key and no daily limit.

Zero retention on any key

Create the key with zero_retention and a search made with it keeps no request text and no answer after the reply. spend_cap_cents is the most the key can spend.

curl -X POST https://api.specstream.ai/v1/account/keys \
  -H "Authorization: Bearer $SPECSTREAM_API_KEY" -H "Content-Type: application/json" \
  -d '{"name": "production", "scopes": ["account:read", "finds:spend"], "spend_cap_cents": 500000, "zero_retention": true}'

Search, pages and rows on one key

The same key reads up to 20 pages or documents a call with the Extract API at $0.80 per 1,000 pages, and a page with no text is not charged. Tavily Extract and Exa Contents calls answer on the same routes. Deep Search takes a request in plain words and returns every match on the live web as structured rows at 5¢, 15¢ or $1 a row. The most a run can cost is approved before it starts.

Brave Search API

$5 per 1,000 requests, with $5 in free credits every month and 50 queries a second. SpecStream answers Brave's GET /res/v1/web/search in Brave's JSON with the same key that answers Tavily and Exa.

Serper

Google results in prepaid packs from $1.00 to $0.30 per 1,000 credits, valid for 6 months. SpecStream answers Serper's call at https://api.specstream.ai/serper/search, in Serper's JSON, from credit that never expires.

Parallel Search API

From $1 per 1,000 requests. SpecStream answers Parallel's POST /v1/search at https://api.specstream.ai/parallel, in Parallel's JSON, on the same key and balance.

Firecrawl

Hobby costs $19 a month, and a search costs 2 credits per 10 results. SpecStream answers Firecrawl's search, scrape, map and crawl calls at https://api.specstream.ai/firecrawl, with search at $5 per 1,000 and pages at $0.80 per 1,000.