Skip to the page
Get started free

OpenAI deep research API in 2026

Shut down, and what to use now.

OpenAI shut down o3-deep-research and o4-mini-deep-research on July 23, 2026. Its deprecations page names gpt-5.6-sol as the replacement: $4 input and $20 output per 1M tokens, plus web search at $10 per 1,000 calls and the tokens it returns.

SpecStream is the live web data engine for teams and agents. Deep Search takes the research request in plain words and returns every match as structured rows, at 5¢, 15¢ or $1 a row, under a maximum you approve before the run.

Get an API key

New accounts get $10 of credit, no card.

SpecStream, the retired deep research models and their replacement

SpecStreamOpenAI deep research (until July 23, 2026)OpenAI gpt-5.6-sol with web search
Status in November 2026Live: Deep Search, the Search API and the Extract API on one keyShut down on July 23, 2026: o3-deep-research and o4-mini-deep-research, with their dated snapshotsThe replacement OpenAI names for both models on its deprecations page
What you sendThe data you want in plain words, with the conditions every row must meetA research prompt to the Responses API, with the required web_search_preview tool and optional code_interpreterA prompt with the web_search tool. return_token_budget: unlimited lets a long research run read more pages
What you get backStructured rows with the columns you asked for, one record per match. Source URL and read time on every valueA citation-rich reportThe model's text answer, written from the search results it read
Price5¢, 15¢ or $1 a delivered row, plus 5¢ for an email and 15¢ for a phone. The searches and pages behind a row are not billed$10 input and $40 output per 1M tokens on o3-deep-research, plus its tool calls$4 input and $20 output per 1M tokens, plus $10 per 1,000 web search calls and the search content tokens at model rates
Price before the runThe most the run can cost, quoted first. Nothing runs until you approve it, and the charge never goes past itMetered by tokens and tool calls as the run wentMetered by tokens and tool calls as the run goes
Long runsA run works until it reaches the count, up to four hours on the largest lists. get_find shows the progressBackground mode, because a task took several minutesBackground mode (background: true) for long multi-search tasks
Search on its ownSearch API: $5 per 1,000 searches of up to 10 results, or $1 with speed=fast. 97.9 F-score on all 1,000 SimpleQA Verified questionsWeb search ran inside the research taskThe web search tool at $10 per 1,000 calls, plus search content tokens

Why teams moved their research to SpecStream

What happened to the OpenAI deep research API

OpenAI shut down o3-deep-research and o4-mini-deep-research on July 23, 2026. A call to either model no longer runs, and OpenAI's cookbook marks its Deep Research API recipe as archived. The deprecations page names gpt-5.6-sol with the web_search tool in their place: a general model that you prompt to research, metered by tokens and by search calls. The guides that still rank for this phrase show the retired models.

Keep the OpenAI SDK, and give it a better search

If your code is built on the Responses API, keep it. Add SpecStream's MCP server at https://api.specstream.ai/mcp as a tool, and the model searches with search_web and reads pages with read_pages. A search costs $5 per 1,000 against $10 per 1,000 for OpenAI's web search tool, and the Extract API reads pages at $0.80 per 1,000. The setup is on the OpenAI API page.

Rows in place of a report

A deep research report answers a question once, and someone still copies its facts into a table. Deep Search returns each match as a row with the columns you asked for: trials, companies, people, filings, tenders or products. Every row meets the conditions you wrote, and get_find_decisions returns each condition's answer with its reason, the value, the source and the time it was read.

Source URL and read time on every value.

The price is set before the run

On o3-deep-research the cost came from the tokens and tool calls a run used, so you learned it after the run. SpecStream quotes the most the whole job can cost, good for 30 minutes. compile reads your request into conditions and columns you can edit, quote_capacity_work prices it, and the run starts only when you approve that maximum as confirm_max_cents. Whatever the run does not use goes back to your balance, and a run that finds nothing costs nothing.

The search under it is measured in public

The SpecStream Search API 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. Scores and method are on the benchmarks page, and what every other deep research API costs per run is on deep research APIs in 2026.

Other replacements for the OpenAI deep research API

Gemini Deep Research

Google's Deep Research and Deep Research Max agents, in preview. Google estimates $1.00 to $3.00 a task, and $3.00 to $7.00 on Max, at preview rates that can change. The agent does not support structured outputs. SpecStream returns structured rows with a source on every value.

Parallel Task API

Nine processors from lite at $5 per 1,000 runs to ultra8x at $2,400 per 1,000. The ultra processor costs $300 per 1,000, or 30¢ a run, and takes 5 to 25 minutes. SpecStream bills the rows delivered, not the runs, so a run that finds nothing costs nothing.

Perplexity Agent API

Support for Sonar Chat Completions ended on September 27, 2026, and research moved to the Agent API with effort levels. The cost varies with the model, the tokens and the tool calls. SpecStream names the most a run can cost before it starts.

Linkup

A search API with a /research endpoint at $0.25 to $2.50 a request, and the first 4,000 queries free. It returns a sourced answer. SpecStream returns every match as a row you can load into a sheet, a CRM or your agent.

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 from o3-deep-research to SpecStream

Sign up and you get $10 of credit, no card. At 5¢ a row, that is up to 200 rows. Create an API key under Agents & API in the workspace.

To keep your OpenAI code, swap the retired model for gpt-5.6-sol and give it SpecStream as an MCP tool. Long runs go in background mode, as before.

import os
from openai import OpenAI

client = OpenAI()

response = client.responses.create(
    model="gpt-5.6-sol",
    background=True,
    tools=[{
        "type": "mcp",
        "server_label": "specstream",
        "server_url": "https://api.specstream.ai/mcp",
        "authorization": os.environ["SPECSTREAM_API_KEY"],
        "allowed_tools": ["search_web", "read_pages"],
        "require_approval": "never",
    }],
    input="Research the economic impact of semaglutide on global healthcare systems.",
)
print(response.id, response.status)

When the research is a list, send it to Deep Search. 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": "Clinical trials of GLP-1 drugs for obesity that started in 2026, with the sponsor, the phase, the number of patients and the registry link", "count": 25}'

Approve that maximum by card through Stripe's Link agent wallet, or in USDC on Base, Arbitrum or Polygon, and you pay for the rows found. On an account, your agent calls compile, quote_capacity_work and authorize_capacity_work over MCP, then reads the rows with list_find_rows.

Get an API key

Frequently asked questions

Something we didn't answer? Talk to a human

Is the OpenAI deep research API still available?
No. OpenAI shut down o3-deep-research and o4-mini-deep-research on July 23, 2026, and a call to either no longer runs. Its deprecations page names gpt-5.6-sol as the replacement, with web search at $10 per 1,000 calls plus tokens.
What replaced o3-deep-research?
gpt-5.6-sol, at $4 input and $20 output per 1M tokens, used with the web_search tool and background mode for long runs. On SpecStream, Deep Search takes the same request in plain words and returns structured rows at 5¢, 15¢ or $1 each.
How much did the OpenAI deep research API cost?
$10 per 1M input tokens and $40 per 1M output tokens on o3-deep-research, plus its tool calls. The cost of a run was known only after it ran. SpecStream quotes the most a run can cost before it starts.
What is the best alternative to the OpenAI deep research API?
SpecStream Deep Search, at 5¢, 15¢ or $1 a delivered row. You describe the data in plain words, approve the most the run can cost, and get every match as a structured row with a source URL and read time on every value. The Search API on the same key scores 97.9 F-score on all 1,000 SimpleQA Verified questions.
Can I keep using the OpenAI SDK?
Yes. Add https://api.specstream.ai/mcp as an MCP tool in the Responses API, with your SpecStream key in authorization. The model then searches at $5 per 1,000 searches and reads pages at $0.80 per 1,000. The steps are on the OpenAI API page.
How long does a Deep Search run take?
Up to four hours on the largest lists, and it works until it reaches the count. get_find shows the progress while it runs. For one fact, the Search API answers in a few seconds from $1 per 1,000 searches.
Does SpecStream have an API and an MCP server?
Yes, and new accounts get $10 of credit, no card. The MCP server is at https://api.specstream.ai/mcp over Streamable HTTP, with OAuth or an API key. There is a REST API, Python and TypeScript SDKs, a CLI and A2A.

Sources

SpecStream wrote this page. Each OpenAI figure comes from OpenAI's own pages, and each other figure from that company's own page.

  1. SpecStream pricingspecstream.ai/pricing, read Nov 12, 2026
  2. SpecStream for agentsspecstream.ai/agents.md, read Nov 12, 2026
  3. SpecStream benchmarksspecstream.ai/benchmarks, read Nov 12, 2026
  4. OpenAI o3-deep-researchdevelopers.openai.com/api/docs/models/o3-deep-research, read Nov 12, 2026
  5. OpenAI o4-mini-deep-researchdevelopers.openai.com/api/docs/models/o4-mini-deep-research, read Nov 12, 2026
  6. OpenAI deprecationsdevelopers.openai.com/api/docs/deprecations, read Nov 12, 2026
  7. OpenAI pricingdevelopers.openai.com/api/docs/pricing, read Nov 12, 2026
  8. OpenAI web search guidedevelopers.openai.com/api/docs/guides/tools-web-search, read Nov 12, 2026
  9. OpenAI cookbook: Introduction to the Deep Research APIdevelopers.openai.com/cookbook/examples/deep_research_api/introduction_to_deep_research_api, read Nov 12, 2026
  10. Gemini Deep Researchaistudio.google.com/docs/deep-research, read Nov 12, 2026
  11. Parallel pricingdocs.parallel.ai/getting-started/pricing.md, read Nov 12, 2026
  12. Perplexity modelsdocs.perplexity.ai/getting-started/models/models/sonar-deep-research, read Nov 12, 2026
  13. Perplexity pricingdocs.perplexity.ai/guides/pricing, read Nov 12, 2026
  14. Linkup pricinglinkup.so/pricing, read Nov 12, 2026