Grounding with Bing Search alternative

The results come back to your code

curl -G https://api.specstream.ai/v7.0/search \
  -H "Ocp-Apim-Subscription-Key: $SPECSTREAM_API_KEY" \
  --data-urlencode "q=Who founded Mistral AI and when?" -d count=10

The reply is Bing's v7 JSON with every result in webPages.value[]. Your code ranks them, reads their pages or hands them to any model, and nothing has to be shown in a fixed form.

Your v7 code needs no rewrite

The route reads q, count (1 to 50, default 10), offset, freshness (Day, Week, Month, a range or one date) and textFormat. mkt, setLang, cc, safeSearch, responseFilter, answerCount, promote and textDecorations are accepted, so a request written for Bing needs no edit. Errors come back as ErrorResponse with Bing's codes. dateLastCrawled is not in the reply: each search reads the live web at the moment of the call.

Moving from the retired API is a one line change, shown on the Bing Search API retirement page.

$5 per 1,000 searches against $14 per 1,000 transactions

A Grounding transaction is one tool call in an agent's run, at $14 per 1,000. A SpecStream search of up to 10 results takes $0.005 from your balance, each started block of 10 more adds $0.006, and a top-up never expires. A model on any provider can call it, and you pay for the search alone.

The most accurate web search API, at about one second

SpecStream scores 97.9 F-score on all 1,000 official SimpleQA Verified questions (978 correct), and 64.0 item F1 and 43.4 row F1 on all 200 official WideSearch tasks. The Extract API scores 0.908 F1 on WCXB. Each score and its method is on the benchmarks page.

On November 9, 2026 production took 1 to 20 searches a second through the fast search every drop-in route uses, and answered 718 of 720. 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.

You decide what is kept

Grounding with Bing Search sends your data outside the Azure compliance and Geo boundary. On SpecStream, zero retention is a field on each key: a search made with it keeps no request text and no answer after the reply. The charge keeps the account, the key, the request ID, the price and the time.

POST https://api.specstream.ai/v1/account/keys

{
  "name": "production",
  "scopes": ["account:read", "finds:spend"],
  "spend_cap_cents": 500000,
  "zero_retention": true
}

Search, pages and rows on one key

GET /v1/web-search returns up to 100 results for one search and an answer with the page it was read from. The Extract API reads up to 20 pages or documents a call at $0.80 per 1,000 pages. Deep Search returns every match on the live web as structured rows at 5¢, 15¢ or $1 a row, with the most a run can cost quoted before it starts. Agents connect to https://api.specstream.ai/mcp.

Grounding with Bing Custom Search

Microsoft's version limited to the sites you choose, at the same $14 per 1,000 transactions, with no access to the raw content either. Its documentation sits under Agents (classic), which Microsoft says will be retired on March 31, 2027. On SpecStream, site: in q keeps a search to one site at $0.005, and the results come back to your code.

Brave Search API

A search API on Brave's own index at $5 per 1,000 requests, with $5 in free credits every month. Its reply has Brave's shape, so Bing code is rewritten for it. SpecStream answers both: Bing's call at https://api.specstream.ai/v7.0/search and Brave's at https://api.specstream.ai/res/v1/web/search.

Tavily

A search API for agents billed in credits, at $0.008 a credit pay as you go. SpecStream answers Tavily's call at https://api.specstream.ai/tavily/search for $0.005 a search.

Exa

A search API priced from $4 per 1,000 requests of up to 10 results, in Exa's own reply shape. SpecStream answers Exa's call at https://api.specstream.ai/exa/search on the same key as Bing's.

Parallel

Search, Extract and other APIs sold by the request. Search starts at $1 per 1,000 requests, and the Build plan has no subscription fee. On SpecStream one request in plain words also returns the finished rows, under a maximum you approve.