Out of deep research tasks? Four moves that work in the next minute
A zero on the deep research counter stops the report, not the work. Four moves, cheapest first:
- Read the counter, then date the refill. OpenAI says the counter inside ChatGPT "shows your remaining tasks" but not where it sits. In May and June 2025, users on OpenAI's community forum reported the hover count had gone. One found the tasks left and the refill date in the dialog that opens when you stop a running report. The refill comes 30 days after the first task of your current window. Find the first deep research report since your last refill in your chat history and add 30 days: a first task on 14 September refills on 14 October.
- Split off the quick facts now. OpenAI's own rule on its deep research help page is "Use search for quick facts, and use deep research for depth and thoroughness." A search answer is not a deep research task, so gather the prices and dates in an ordinary chat today and keep the next task for the synthesis.
- Run the report on another vendor's meter. Gemini Deep Research is open to Gemini users with or without a Google AI plan (subject to demand), Perplexity's free plan includes 1 Research query a month, and Claude Research comes with any paid Claude plan. Paste the same brief; each counts against its own vendor's limit, never against OpenAI's.
- Do not upgrade on a guess. On 3 October 2026 ChatGPT Pro starts at $100 a month and its pricing card promises "Maximum deep research" with no number next to it, so count how often you hit zero over two windows before paying for a bigger counter you cannot see in advance.
How many deep research tasks does each ChatGPT plan get?
No ChatGPT plan has a published deep research number any more. On 3 October 2026 the ChatGPT pricing page describes the allowance in words only. The last counts OpenAI printed are from 2025, when the FAQ split each plan into full and lightweight tasks:
| Plan | What chatgpt.com/pricing says, 3 October 2026 | Last count OpenAI published (help page, June 2025 version) |
|---|---|---|
| Free | "Limited deep research" | 2025: 5 tasks a month, all lightweight |
| Go | "Limited" in the plan comparison table | No count on any OpenAI page we read |
| Plus | "Expanded deep research" | 2025: 25 a month, 10 full plus 15 lightweight |
| Pro (from $100 a month) | "Maximum deep research" | 2025: 250 a month, 125 full plus 125 lightweight |
| Team, Enterprise and Edu (2025 names) | Business and Enterprise sit on a separate pricing tab | 2025: 25 a month, 10 full plus 15 lightweight |
We compared Wayback Machine copies of OpenAI's deep research FAQ to date the change. The 16 July 2025 copy still lists every plan's count. The 22 August 2025 copy replaces them with "Deep Research usage varies by plan". Every copy since, through April 2026, keeps that line. The 25, 250 and 5 that most pages still quote come from the 24 April 2025 note on OpenAI's Introducing deep research post. It's still there. Only your counter can say whether today's allowance matches them. Treat those numbers as the floor OpenAI once committed to, not as a current promise.
When does the ChatGPT deep research limit reset?
The ChatGPT deep research limit resets 30 days after your first task, on a rolling window of your own, not on the 1st of the month and not on your billing date. OpenAI's help page puts it in one sentence: "For plans with a fixed monthly allowance, it resets every 30 days from the date of your first use." The clock starts with a task, not with the subscription. Two people who joined Plus on the same day can refill a week apart.
The June 2025 version of the FAQ spelled out two details the current page trims. The first task "begins" a 30 day usage window that then repeats from that point. That window "carries over even after switching to a different plan". So moving from Plus to Pro in the middle of a window didn't hand you a fresh one. Nothing on the April 2026 copy of the page says that changed. If the counter hits zero on day 12, you're 18 days from a refill. Or you switch tools.
What is the difference between full and lightweight deep research?
Lightweight deep research is the cheaper version OpenAI added on 24 April 2025, "powered by a version of o4-mini". ChatGPT moved to it by itself once a plan's full tasks were used up. In the 2025 counts it was most of the allowance on Plus (15 of 25), half on Pro (125 of 250) and all of it on Free (5 of 5). OpenAI described it as "designed to be more cost-efficient while preserving high quality", on a smaller model than the full version.
The split has since dropped out of OpenAI's documentation. The help page's April 2026 copy says deep research "is powered by the latest models", with an option to pick a legacy model for a task. It no longer mentions a lightweight tier or when a switch happens. So a report that reads thinner than last month's isn't proof you were moved to lightweight. OpenAI doesn't say either way any more. Agent mode is a separate door. Since 17 July 2025 OpenAI says deep research "can now go even deeper and broader" there, with a visual browser. The original tool stays in the + menu, at /Deepresearch and in the sidebar.
How do Gemini, Claude and Perplexity cap deep research?
Only Perplexity prints a count, 1 a month on Free. All three describe their paid tiers in words. We read each vendor's own help pages on 3 October 2026, for the consumer plans:
| Tool | Who gets it | Count published for the main paid plan | When it refills |
|---|---|---|---|
| ChatGPT deep research | Free and Go (Limited), Plus, Pro | None; Plus is "Expanded" | 30 days from your first task |
| Gemini Deep Research | Gemini users with or without a plan; AI Pro and Ultra get "higher limits" | None; the app tells you how many requests are left for the day | Daily request limit, inside a usage meter that refreshes every 5 hours up to a weekly cap |
| Claude Research | Pro, Max, Team and Enterprise | None; "subject to the same limits as standard Claude conversations" | Session limit every five hours, plus a weekly limit at a fixed time |
| Perplexity Research | Free (1 a month), Pro, Max | None; Pro reads "Monthly limits (average use)" | Monthly |
| Whizi Deep Research | Paid plans, in the web app | 200 credits a run, so Pro's 2,500 monthly credits cover 12 | When the UTC calendar month turns over |
Try Gemini first when ChatGPT runs dry. It runs without a paid plan at all. Google's Deep Research help page caps both "Daily research requests" and the number you can run at once. Its limits page adds that Deep Research "will consume more of your usage" on a meter that "refreshes every 5 hours until you reach your weekly limit". AI Pro gets 4x the standard allowance. Claude Research has no separate counter. It does warn that research sessions "can use up your limits faster". Perplexity's plan table prints 1 a month for Free and words for Pro. Its April 2026 Advanced Deep Research note, still up on 3 October 2026, says "Usage limits have been adjusted to allocate more computing power per session". That reads as fewer, heavier runs. The Perplexity Pro limits guide covers the rest of Pro's caps.
Whizi, which publishes this page, prices its Deep Research in credits rather than tasks. A run costs 200 credits, and the request can be 8 to 2,000 characters. A run that fails at the planning or search stage ends with "You have not been charged." Where it loses: Pro's 2,500 credits at $29.99 a month buy 12 runs if spent on nothing else. That's fewer than the 25 OpenAI last published for Plus at $20. Only one run can go at a time. And every run comes out of the same credits as your chats. What you get instead is a fixed price per run that you know before you start. The refill comes on a date everyone shares.
- Read the deep research counter in ChatGPT before anything else; it is the only number OpenAI shows for your plan
- Date your refill as the first deep research task of the current window plus 30 days, not the 1st and not your billing date
- Move quick lookups to ordinary chat with search so the next deep research task goes on synthesis
- If the report cannot wait, paste the same brief into Gemini Deep Research, Perplexity Research or Claude Research, each on its own meter
- Count how often you hit zero over two 30 day windows before paying from $100 a month for Pro's unnumbered "Maximum"
Frequently asked questions
How many deep research tasks does ChatGPT Plus get?
OpenAI last published 25 a month for Plus: 10 full tasks plus 15 lightweight ones, in its April 2025 update and in its help page FAQ until August 2025. Since August 2025 the help page says only that usage varies by plan. As of October 2026 the Plus card on chatgpt.com/pricing reads "Expanded deep research" with no number. The counter inside ChatGPT shows what you have left.
Does ChatGPT Go include deep research?
Yes, at a limited level. As of October 2026 the ChatGPT pricing page's comparison table marks deep research as "Limited" on both Free and Go and "Yes" on Plus and Pro. OpenAI labels both Free and Go Limited and publishes no count for either. Nothing on its pages says whether Go gets more runs than Free.
How do I check how many deep research tasks I have left?
Use the usage counter inside ChatGPT, which OpenAI says "shows your remaining tasks". It's the only current figure, because OpenAI stopped printing per plan counts in August 2025. For the refill date, add 30 days to the first deep research task of your current window.
Does upgrading to Pro reset the deep research limit?
No, going by the last version of OpenAI's FAQ that addressed it: in June 2025 it said the 30 day usage window "carries over even after switching to a different plan". An upgrade raises the ceiling inside the window you're already in. It doesn't start a new 30 days.
