DeepSeek token counter & API cost calculator
Paste any text to see how many tokens DeepSeek reads it as, and what it costs on each model. Counts update as you type.
| Model | Input $/M | Output $/M | Context | Per call | Per month |
|---|---|---|---|---|---|
|
DeepSeek V4 Pro
DeepSeek |
$1.32 | $3.96 | 1M | — | — |
|
DeepSeek V4 Flash
DeepSeek 3 hosts · up to ×6.3 dearer elsewhere |
$0.44 | $1.32 | 1M | — | — |
Rates are USD per million tokens, published list price, checked 2026-08-31. Batch and cached-input discounts are not applied. Confirm against the provider before committing a budget.
How DeepSeek counts tokens
DeepSeek uses its own tokeniser vocabulary. This calculator does not model it separately — it applies one generic multiplier to every provider outside the big three, so treat the DeepSeek figure as a close approximation rather than an exact count.
The counter above is an estimate, not DeepSeek's own merge table. It models the behaviour that actually drives the number — a leading space merging into the word after it, digits grouping, characters outside the Latin range costing several times more — which is close enough to size a prompt and budget a workload, and deliberately not close enough to reconcile an invoice against.
DeepSeek is the only major provider that charges by time of day
Every other vendor on this site charges the same rate at three in the morning as at three in the afternoon. DeepSeek does not: it publishes an off-peak window at half the peak rate. That is not a discount you negotiate or a tier you qualify for — it is the published price, and it applies to anyone whose work can wait. For batch jobs, overnight summarisation, backfills and anything else with no user sitting in front of it, that halves the bill for no change other than when the job runs. The table shows the peak rate, so read it as the ceiling rather than the price.
DeepSeek V4 Flash: Peak rate. Half off-peak, and third-party hosts run it cheaper still - see the range.
DeepSeek V4 Pro: Peak rate. DeepSeek bills half this off-peak, so the same job can cost half as much overnight.
What the 2 DeepSeek models cost
Take one representative call — a 2,000-token prompt and a 600-token reply, which is roughly two pages in and one page out — and run it 100 times a day.
the cheapest here
the dearest here
for identical work
on the dearest, vs $5.02 on the cheapest
DeepSeek is already among the cheapest frontier-class options at peak, and the off-peak window puts it in a bracket of its own. The catch is that the same model is also served by third-party hosts at rates that can undercut even the off-peak price, so if cost is the deciding factor, check the host range before assuming first-party is cheapest.
Which DeepSeek model to use
Flash for volume, Pro when the task genuinely needs the depth. The gap between them is roughly threefold, and Flash handles summarising, extraction and routine drafting well enough that Pro is a deliberate choice rather than a default. Both carry a very large context window, so the decision is about reasoning depth rather than how much you can fit in.
Context windows
The DeepSeek models here run from 1M tokens to 1M. That window is shared: your prompt, every earlier turn in the conversation, and the reply all have to fit inside it together. The table above marks a model in red when the text you have pasted plus the reply length you set would not fit — which is usually how people discover that a window is not as generous as the headline number suggests.
Questions
What is DeepSeek off-peak pricing?
DeepSeek publishes a discounted rate during a defined off-peak window, at roughly half the peak rate for both input and output. It is the standard published price rather than a promotion, so any workload that can be scheduled overnight pays about half. The table on this page shows peak rates.
Is DeepSeek cheaper than GPT or Claude?
On published rates, considerably — and off-peak, dramatically. Whether it is cheaper for your workload depends on how many attempts you need to get a usable answer, which no pricing table can tell you. Price the job, not the token.
Is this DeepSeek official token count?
No. This is an estimate using a generic multiplier, because DeepSeek tokenisers are not separately modelled here. It is close enough for budgeting and not close enough for invoice reconciliation. DeepSeek documentation is the authority on an exact count.
Can I use DeepSeek models through another host?
Yes, and often more cheaply. Open-weight DeepSeek models are served by several third-party hosts at rates that can undercut the first-party API. Where we hold real rates for more than one host, the table marks the spread against that model.
Counting for a different provider
The same text costs a different number of tokens on every family, so if you are comparing providers, compare on your own text rather than on a rule of thumb.
Rates are USD per million tokens, published list price, checked 2026-08-31. Batch and cached-input discounts are not applied. Confirm against DeepSeek before committing a budget.
More free tools
All free, all instant, none of them need an account.