Skip to the page
Get started free

Sonar Deep Research after Sept 27, 2026

Move it to a cited report from $0.10.

Perplexity ended Sonar Chat Completions on September 27, 2026. Synchronous sonar-deep-research calls are reformulated as Agent API requests, and asynchronous ones are no longer supported. Perplexity maps sonar-deep-research to the high effort level, billed in tokens plus each tool call, so the price of a run is known after it ends.

SpecStream is the live web data engine for teams and agents. POST /v1/research returns a written report with the page, the read time and the passage behind each cited sentence, at one flat price per depth: $0.10 quick, $0.30 standard or $1.00 deep. It runs in the background by design: 202 at once, then poll, stream or a webhook.

Get an API key

New accounts get $10 of credit, no card. It covers 100 quick reports, 33 standard reports or 10 deep reports.

A deep research call on SpecStream and on Perplexity's Agent API

SpecStreamPerplexity Agent API
Price of one run$0.10 quick, $0.30 standard, $1.00 deep. One flat price per depthTokens at the model's rates plus each tool call, such as $2.50 per 1,000 web searches and $0.0005 per page fetch
When you know the priceBefore the run. The reply to the POST shows price_usdAfter the run. Perplexity's pricing page says costs vary with the model, token usage and tool invocations
Cost per request, as published$0.10, $0.30 or $1.00, the same on every questionOn Perplexity's own benchmark page: high $0.1360 on BrowseComp and $0.1210 on DSQA, xhigh $0.7800 and $0.2180. Sonar Deep Research was $0.6110 and $0.5656
Long runsEvery report runs in the background: 202 at once, then GET /v1/research/{id}, server-sent events on /events, or a webhookBackground mode: submit with background: true and poll the response by id. Async Sonar requests stopped
Citationsreport.citations gives each cited sentence its page, its read time and its passage. report.sources lists every page readInline citations for source-backed claims, source typed as [web:1] on low, medium and high
JSON in place of text"output": {"type": "json", "schema": {...}} with your JSON SchemaStructured outputs with a JSON Schema
Search and pages on the same keySearch API at $5 per 1,000 searches and $1 on the fast tier, Extract API at $0.80 per 1,000 pages, and a drop-in address for your Perplexity Search API codeSearch API at $5 per 1,000 requests, or $1 with Fast Search

Why Sonar Deep Research users move to SpecStream

A fixed price per report, where Sonar had a token bill

sonar-deep-research was billed in tokens, with its citation and reasoning tokens charged, and the Agent API keeps a bill of the same kind: tokens at the model's rates plus each tool call. Perplexity's own table puts a high run at about $0.12 to $0.14 and an xhigh run at $0.22 to $0.78, depending on the questions.

A SpecStream report costs $0.10, $0.30 or $1.00 by depth, on every question. The reply to the POST shows price_usd before anything is charged, the price is taken once when the report is delivered, and a cancelled report costs nothing. A product that resells answers can quote its own price on that.

Your async loop keeps its shape

The async Sonar API is the part that broke on September 27. A SpecStream report is asynchronous by design, so the loop you already have keeps its shape: submit, keep the id, read until status is completed. GET /v1/research/{id}/events streams progress as server-sent events, and Last-Event-ID resumes a dropped stream. A webhook is called when the report ends, so a worker does not poll.

Each cited sentence shows its page and when it was read

report.content holds the text with [1] markers. report.citations gives each cited sentence the URL of its page, the time SpecStream read it and the passage that supports it, and report.sources lists every page read. A reviewer checks a claim in seconds, without opening every source.

Search, pages, rows and reports on one key

The same key runs the Search API, which scores 97.9 F-score on all 1,000 SimpleQA Verified questions, and the Extract API at $0.80 per 1,000 pages. Perplexity Search API code runs on SpecStream when you set the base address to https://api.specstream.ai/perplexity. When the answer is a table, Deep Search returns every match as a row. See the benchmarks, research reports and pricing.

Other Sonar Deep Research alternatives

Tavily Research

Tavily Research bills 15 to 250 credits a request on pro, at $0.008 a credit pay as you go: $0.12 to $2.00, known after the run. A SpecStream standard report costs $0.30 every time.

Linkup Research

Linkup prices Research at $0.25 to $2.50 a request by reasoning depth, in 1 to 10 minutes. A SpecStream quick report costs $0.10 and takes about a minute.

Parallel Task API

Parallel's deep research runs on pro at $100 per 1,000 runs, in 2 to 10 minutes, or on ultra at $300 per 1,000, in 5 to 25 minutes. SpecStream returns a cited report in about one to five minutes, with the passage behind each cited sentence.

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 move a sonar-deep-research call, step by step

1. Create a key under Agents & API in the workspace. A new account gets $10 of credit, no card.

2. Send the user's question as input to POST /v1/research, where Sonar took messages. Put your system prompt at the top of the same input: it holds up to 15,000 characters. Choose depth where you set model and reasoning_effort: quick for low, standard for medium, deep for high.

curl -X POST https://api.specstream.ai/v1/research \
  -H "Authorization: Bearer $SPECSTREAM_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: report-0001" \
  -d '{"input": "Compare the 2024 cloud revenue growth of Microsoft, Amazon and Google.", "depth": "deep"}'

# 202 at once, with the report's id and price_usd. Then read it:
curl https://api.specstream.ai/v1/research/<id> \
  -H "Authorization: Bearer $SPECSTREAM_API_KEY"

3. Where your code read choices[0].message.content, read report.content. Where it read citations or search_results, read report.citations for each cited sentence and report.sources for every page read.

4. For response_format with a JSON Schema, send "output": {"type": "json", "schema": {...}}. search_domain_filter becomes include_domains, and a domain you excluded goes in exclude_domains. Add "webhook": {"url": "https://..."} to be called when the report ends.

From an agent, connect the MCP server at https://api.specstream.ai/mcp: research_price shows the price of each depth, start_research starts the report once your person approves it, and get_research reads it. The research reports guide has every field. For Perplexity Search API calls, the Perplexity migration guide keeps your code at $1 per 1,000 searches.

Get an API key

Frequently asked questions

Something we didn't answer? Talk to a human

Is sonar-deep-research still available?
Not on Sonar Chat Completions: Perplexity ended its support on September 27, 2026. Synchronous calls are being reformulated as Agent API requests, rolling out by model, and asynchronous Sonar requests are no longer supported.
What replaces sonar-deep-research in Perplexity's API?
The Agent API at https://api.perplexity.ai/v1/agent. Perplexity's migration guide maps sonar-deep-research to the high effort level and points to xhigh for its strongest deep research. The ultra effort level, added in November 2026, is for broad research where coverage matters most.
What does a deep research run cost on the Agent API?
Tokens at the model's rates plus each tool call, so the price varies by question. On Perplexity's own benchmark page, high cost $0.1360 a request on BrowseComp and $0.1210 on DSQA, and xhigh cost $0.7800 and $0.2180.
What is the best Sonar Deep Research alternative?
SpecStream, at one flat price per report: $0.10 quick, $0.30 standard or $1.00 deep. Each cited sentence carries the URL of its page, the time it was read and the passage, and the call runs in the background with progress events and a webhook.
How long does a SpecStream research report take?
About one minute for a quick report, four minutes for a standard report and five minutes for a deep report.
What does a cancelled report cost on SpecStream?
$0. The price is taken once, when the report is delivered, and a cancelled report is not charged.
Can I keep my Perplexity Search API code?
Yes. Set the base address to https://api.specstream.ai/perplexity and POST /perplexity/search answers in Perplexity's shape, at $1 per 1,000 searches of up to 10 results.

Sources

SpecStream wrote this page. Each figure about another provider comes from that company's own page, read on the date shown.

  1. Perplexity: Migrating from Sonardocs.perplexity.ai/docs/agent-api/migrate-from-sonar/overview, read Nov 13, 2026
  2. Perplexity: migration how-todocs.perplexity.ai/docs/agent-api/migrate-from-sonar/how-to.md, read Nov 13, 2026
  3. Perplexity: Agent API vs Sonar benchmarksdocs.perplexity.ai/docs/agent-api/migrate-from-sonar/benchmarks.md, read Nov 13, 2026
  4. Perplexity pricingdocs.perplexity.ai/docs/getting-started/pricing, read Nov 13, 2026
  5. Perplexity changelogdocs.perplexity.ai/docs/resources/changelog.md, read Nov 13, 2026
  6. Tavily API creditsdocs.tavily.com/documentation/api-credits.md, read Nov 13, 2026
  7. Linkup pricinglinkup.so/pricing, read Nov 13, 2026
  8. Parallel pricingdocs.parallel.ai/getting-started/pricing.md, read Nov 13, 2026
  9. SpecStream research reportsspecstream.ai/docs/research, read Nov 13, 2026
  10. SpecStream pricingspecstream.ai/pricing, read Nov 13, 2026
  11. SpecStream benchmarksspecstream.ai/benchmarks, read Nov 13, 2026