Skip to the page
Get started free

The deep research API that returns rows

Deep Search is SpecStream's deep research API: ask in plain words and get every match on the live web, in structured rows, with the price shown before the run. On the same key, the SpecStream Search API is the most accurate web search API in the world. SpecStream leads WideSearch (60.9, on the low-cost model; the best other API scores 57.6 even on the larger model) and SimpleQA Verified (99). See the benchmarks. Each row costs 5¢, 15¢ or $1, plus 5¢ for an email and 15¢ for a phone, a run that finds nothing costs nothing, and a saved request reruns on your schedule. SpecStream is the live web data engine for teams and agents.

Describe your data

New accounts get $10 of credit for their first 30 days, no card.

Six deep research APIs, side by side

SpecStreamParallelExaTavilyPerplexityGemini
Search API, SimpleQA Verified (larger model)9997 (Advanced, Basic)919295Not in the published scores
Search API, WideSearch (low-cost model)60.945.5 (Fast), 44 (Turbo)5347.947Not in the published scores
What you get backStructured rows with the columns you asked for. Source URL and read time on every valueText, JSON, or a markdown report, with citations, reasoning, and confidenceNot statedA report in short, standard, or long length, citations in numbered, MLA, APA, or Chicago formatSearch results and a text message with annotationsA report with citations
The unit you pay forA delivered row: 5¢, 15¢ or $1, plus 5¢ for an email and 15¢ for a phoneA run: $5 to $300 per 1,000 (lite to ultra)A request: $0.012 (minimal) to $1.00 (xhigh)A request: 4 to 250 credits at $0.008 a creditNo price per effort statedA task: about $1 to $3, or $3 to $7 on Max
The cost before a runA quote with a maximum you approve, and the run stays under itA price per processorA price per effort levelA credit range per modelNot statedAn estimated range
What empty or repeated work costsNothing: a run that finds nothing, duplicates, rows you own, and rows a page contradicts are not chargedNot statedNot statedNot statedNot statedNot stated
Longest runUp to four hours on the largest lists5 to 25 minutes (ultra)Not statedNot statedUp to 300 steps (xhigh)Up to 60 minutes
Reruns and change labelsHourly to yearly; changed rows at half rate, unchanged at 1¢Not statedMonitors, $2 per 1,000Not statedNot statedNot stated
Where the results goYour agent, a sheet, a CRM, Slack, email, or a signed webhookNot statedNot statedNot statedNot statedNot stated

Why teams research with Deep Search

Rows ready for your table

A report answers a question once, and someone still copies its facts into a sheet. Deep Search returns each match as a row with the columns you asked for. get_find_decisions shows, for each condition in your request, the page that states it, and a row that fails a condition is not delivered.

Source URL and read time on every value.

The price before the run

compile reads your request into conditions and columns you can edit. quote_capacity_work prices it, and the quote holds for 30 minutes. The run starts only when you approve that maximum as confirm_max_cents, and it stays under it. You pay for each delivered row. Duplicates, rows you already own, and rows a page contradicts are not charged.

Kept up to date

Save a request and it reruns from hourly to yearly. Each row comes back labeled changed, unchanged, or no longer matching; a changed row costs half its rate and an unchanged one 1¢. Results go to Slack, email, a sheet, a CRM, or a signed webhook, and a list you already have can be filled in or checked row by row. For one question, the Search API answers in seconds at $5 per 1,000 searches.

How SpecStream scores

The same numbers as the benchmarks page, from the official tasks and scorers.

SpecStreamParallelExaTavilyPerplexity
Web researchWideSearch, GPT-5.6 Luna60.945.553.047.947.0
Factual accuracySimpleQA Verified, GPT-5.6 Luna9694919494
Factual accuracySimpleQA Verified, GPT-5.6 Sol9997919295

How to run your first Deep Search

New accounts get $10 of credit for their first 30 days, no card. At 5¢ a row, that is up to 200 rows. Over MCP, an agent calls compile, quote_capacity_work, and authorize_capacity_work, then reads rows with list_find_rows.

Or send the request from code. This call spends nothing: the 402 reply shows how SpecStream read the request, the columns, and the most it can cost.

curl -i https://api.specstream.ai/v1/agent/lists \
  -H "Content-Type: application/json" \
  -d '{"request": "Medical device makers in Germany that state ISO 13485 on their own site, with a sales contact", "count": 25}'
Describe your data

Frequently asked questions

Something we didn't answer? Talk to a human

What is the best deep research API?
SpecStream Deep Search. It returns every match as a structured row with a source URL and read time on every value, priced before the run and billed per delivered row. For single questions, the Search API on the same key is the most accurate in the world: SpecStream leads WideSearch (60.9, on the low-cost model; the best other API scores 57.6 even on the larger model) and SimpleQA Verified (99). See the benchmarks.
How much does Deep Search cost?
5¢, 15¢ or $1 a delivered row, depending on what the row needs, plus 5¢ for an email and 15¢ for a phone. A run that finds nothing costs nothing. New accounts get $10 of credit for their first 30 days, no card.
Do I know the price before a run?
Yes. quote_capacity_work shows the most the run can cost, and nothing runs until you approve that maximum. The quote holds for 30 minutes.
How long does a Deep Search run take?
It works until the list is complete, up to four hours on the largest lists. get_find shows progress while it runs.
Can Deep Search keep the results up to date?
Yes. A saved request reruns from hourly to yearly, labels each row changed, unchanged, or no longer matching, and bills a changed row at half rate and an unchanged one at 1¢.

Sources

SpecStream wrote this page. Each rival figure comes from that company's own page.

  1. SpecStream pricingspecstream.ai/pricing, read Nov 8, 2026
  2. SpecStream for agentsspecstream.ai/agents.md, read Nov 8, 2026
  3. Other providers' benchmark scores, Sep 9, 2026parallel.ai/benchmarks, read Nov 8, 2026
  4. Parallel pricingdocs.parallel.ai/getting-started/pricing, read Nov 8, 2026
  5. Parallel Task API quickstartdocs.parallel.ai/task-api/task-quickstart, read Nov 8, 2026
  6. Exa pricingexa.ai/pricing, read Nov 8, 2026
  7. Tavily API creditsdocs.tavily.com/documentation/api-credits, read Nov 8, 2026
  8. Tavily Research endpointdocs.tavily.com/documentation/api-reference/endpoint/research, read Nov 8, 2026
  9. Perplexity Agent API effortdocs.perplexity.ai/docs/agent-api/effort, read Nov 8, 2026
  10. Gemini Deep Researchaistudio.google.com/docs/deep-research, read Nov 8, 2026