Else vs ChatGPT (2026) — an honest comparison

The short answer. You cannot work it out in dollars, and that is the whole difference. ChatGPT sells a seat by the month — $20 for Plus, $100 and $200 for Pro, $25 per user for Business — and counts what you use in credits: a message costs a set number of credits depending on which model you picked, agent mode costs more per message, a deep-research task costs more still. OpenAI publishes those credit rates. What it does not publish anywhere we could find is what a credit is worth in dollars. Else sells no seat: you top up a balance in dollars, and each run ends with a receipt showing the dollar amount beside the engine tier that produced it. Almost nobody should replace ChatGPT with Else — it does far more things, and it does them on your phone. Read the section on where Else falls short before you decide anything.

The difference in one line

ChatGPT tells you a task took credits.
Else tells you it cost $0.0819 on the Auto tier.

Credit for credit where it is due: OpenAI does tie its unit to the named model you picked, which is more than an unexplained number. The gap is that the unit is not money, and no rate exists on its pages to turn it into money. So you can know which model ran and how many credits it took — and still not know what it cost.

Counted in credits

30 credits

The published rate for one message in agent mode. A deep-research task is 50. Real numbers, honestly disclosed — in a unit with no published dollar value, so the arithmetic stops here.

Counted in dollars

$0.0819

Auto tier · 22 s · real run, 30 July 2026

The price follows the tier you picked, which is why an amount never appears on this site without the tier beside it. Five real receipts, unedited →

What Else does differently

Three things, all checkable today.

This list is short on purpose. ChatGPT does more than Else does; these are the places where the difference is real.

01The unit is the dollar, with the tier attachedNot a credit, not a message allowance. Each run ends with an itemized receipt in US dollars: the total, the limit that was set, how much of it was used, and the lines under it. You pick the tier — Auto, Fast, Smart or Max — or point one request at a named model, and because the price follows that choice, no amount on this site is printed without the tier beside it.

02Nothing monthly to carryNo seat, no plan, no renewal date. You top up a balance, it never expires, and it only moves when something actually runs. Four jobs in a month costs you four jobs. Five real receipts are published unedited, failures included.

03A dollar ceiling on handed-off workWork you hand off gets a spending limit in dollars before it starts — $0.15 by default, up to $5 — and going over it makes the run stop and check with you. On a single chat message that same field is not applied yet: a fixed $0.50-per-turn ceiling stops it instead. Read the honest section below.

Side by side · Data as of 30 July 2026

The dimensions that change a decision.

Every ChatGPT cell is quoted from a page OpenAI publishes; every Else cell is checkable on this site or in the product. Sources and capture dates are filed in the repository. Read every row as true on the date above, not for ever — this is the fastest-moving product on the list.

What you are comparing	Else	ChatGPT
What you pay up front	Nothing. You top up a balance in US dollars	A monthly seat — free, $20 for Plus, $100 and $200 for Pro, $25 per user for Business (or $20 per user billed annually, minimum two seats)
What usage is counted in	US dollars	Credits, at a published rate per message that depends on the model you picked
What a credit or a run is worth in money	It is money	No credit-to-dollar rate is published on the pages we checked
What you know after a run finishes	An itemized receipt in dollars, beside the tier that ran	The credits it consumed
Spending cap you can set	Handed-off work only, in dollars — $0.15 by default, up to $5, and checked after the money is spent. On a single message the field is not applied yet	None found on the personal plans. Business has monthly usage limits, counted in credits, off by default
Choosing the engine	Four named tiers, or one named model for a single request	Models are named and selectable, and which ones you get depends on your plan
Running work on a schedule	Any handed-off job can repeat, from every 30 minutes to weekly	One-off and recurring scheduled tasks, plus an agent mode and change-watching
Team and enterprise accounts	None	Business and Enterprise plans
Public API	None	Yes — and priced per token in dollars, though billed separately from any ChatGPT plan
Apps	Browser only, installable as a web app	Web, iOS, Android, macOS and Windows
Breadth of what it does	Handed-off work, and little else	Chat, work and code in one place, by its own description

The dollar figures for ChatGPT plans are taken from OpenAI's own help articles, which state US dollars, and not from its pricing page — that page serves local currency, and from where we checked it rendered Singapore dollars. The ChatGPT Go tier is left out of the table on purpose: its help article publishes no figure, so we have no verified US-dollar price for it. Where a cell reads “not published” or “none found”, that means it was absent from the pages we fetched on 30 July 2026 — it is a statement about what is published, not about what exists.

Where Else falls short

Four reasons to keep ChatGPT.

Not hedging — these are measured, and two of them are on our own site already.

It does one thing, and ChatGPT does dozens

Else takes a job, does it, and hands back a file with a receipt. ChatGPT is a place to chat, work and code all in one — in your pocket, on your desktop, free if you want it to be, with a model lineup and a feature surface we are not going to pretend to match. If you have one subscription for everything, keeping it is the sane default. The honest question is not which one, it is whether the jobs you hand off deserve to be priced in money.

Open-web research is where we fail

On 30 July 2026 we ran five jobs on the live product. The two that had to read their way around the open web produced no file and no answer — and were charged $0.6169 and $0.6027 anyway, both on the Auto tier. One of them we had deliberately written to be small. ChatGPT has a deep-research mode with a published rate and a long track record. On this job type today, ours is the weaker product.

The spending limit you set is not applied yet

On a single message, the limit you type is sent to us and nothing on our side reads it. What stops the run is a different ceiling you cannot see or change: a fixed $0.50 per turn, checked only after a step has already been billed — so a turn ends at $0.50 plus whatever the step that crossed it cost. On those same five runs the limit was set to $0.15 and four receipts came in between $0.502 and $0.6169, all four on the Auto tier: a tight band just above $0.50, unrelated to the $0.15 we set. On work you hand off the limit is read — but only between steps, on money already spent, and not at all when a run finishes in one go. Showing you dollars is only half of the promise; holding the ceiling is the other half, and we have not built it yet.

Nothing for teams, developers or phones

There is no team account, no shared workspace and no view of a colleague's work — everyone runs from their own account and their own balance. There is no public API. There is no App Store or Google Play app; Else runs in a browser and can be installed as a web app. ChatGPT has all three. And a run that fails here is still billed for the work that actually ran — there is no refund.

When to use each

Most people should use both.

Reach for ChatGPT when…

You want one thing that does everything — chatting, writing, coding, images, voice — on every device you own. Several people need to work in the same place. You are building on an API. You want a free tier, or a single predictable monthly number and no per-job arithmetic. Or the job is wide open-web research, which it prices openly and has been doing for a long time.

Reach for Else when…

You want to see, in dollars, what a particular job cost — and which engine tier that price came from. You hand off a handful of real jobs a month and would rather pay for those than carry a seat. You want a dollar ceiling on a run and to be asked before anything is sent or published. Or you want a job to keep repeating without you and to account for itself each time.

Questions people actually ask

The six we hear most.

How much does ChatGPT agent mode cost?

In credits, OpenAI publishes the rate: one message in agent mode is 30 credits, and a deep-research task is 50. In dollars, you cannot work it out — we could find no credit-to-dollar rate published anywhere on its pages, so any dollar figure would be someone's guess. On the money question, that is the whole difference between the two products: Else charges in dollars and prints the amount on a receipt beside the engine tier that produced it.

Does ChatGPT agent mode cost more?

Yes, in credits, and OpenAI says so plainly: its rate card charges more credits for a message in agent mode than for an ordinary message, and more still for a deep-research task. It also notes that asking for more reasoning effort does not change the per-message rate. What the rate card does not give is a dollar value for a credit.

Can ChatGPT do scheduled tasks?

Yes. OpenAI's own help centre says users can schedule one-off and recurring tasks, and can ask ChatGPT to watch for changes and notify them when something meaningful moves. We are not going to pretend otherwise: this is not a gap. Else can also give any handed-off job a repeat, from every 30 minutes to weekly, and each run of it produces its own receipt in dollars.

Is there a limit on ChatGPT's scheduled tasks?

That is a question for OpenAI's help centre, and we have not verified an answer, so we are not going to publish one. What we can tell you about Else is where its limits are: a spending limit in dollars on handed-off work, $0.15 by default and up to $5, and no weekly or monthly ceiling — so with a frequent repeat, the per-run limit is the number that matters. On a single chat message that field is not applied at all today; a fixed $0.50-per-turn ceiling stops the run instead. Both facts are documented on this page and on our examples page.

What does ChatGPT cost per message?

OpenAI publishes a per-message rate in credits that depends on which model you picked — 10 credits on one model and 50 on a higher one, on the rate card for its business plans. It does not publish what a credit costs in money. Else has no per-message price at all: you are charged for what a run actually used, in dollars, and the receipt shows the total beside the tier it ran on.

How much does an AI agent cost?

For Else we published five real runs from one afternoon rather than an average. On the Auto tier they came in at $0.0819, $0.502, $0.5102, $0.6027 and $0.6169. Read that honestly: three produced a file you can open, and two produced nothing at all and were still charged. A different tier gives a different receipt, which is why we never quote an Else price without naming the tier.

ChatGPT tells you a task took credits.Else tells you it cost $0.0819 on the Auto tier.