Tokens, Seats, Credits and Minutes: A Practical Dictionary of AI Billing Units
Short answer. AI tools use different billing units because they charge for different aspects of access and consumption. Tokens generally measure text processing, seats count users, credits represent provider-defined allowances and minutes measure time-based services. Understanding the unit behind a listed price is essential before comparing subscriptions or estimating an AI subscription cost.
- A subscription price and a usage price describe different parts of an AI service.
- Tokens measure processed text, not necessarily words or characters.
- Seats count authorised users, while credits represent an allowance defined by the provider.
- Minutes can measure audio processing, video duration or other time-based consumption.
- A useful AI tool pricing comparison starts with the billing unit and the included allowance.
Why the billing unit matters more than the headline price
An AI pricing page can display a monthly subscription, a price per million tokens, a number of credits or a charge per minute. These figures describe different purchasing arrangements. A monthly subscription may provide access to a collection of features, while a usage-based charge may increase with every request. A price comparison that ignores this distinction can make two services appear comparable when their commercial structures are fundamentally different.
Start by identifying what the customer is purchasing. Is it access for one person, permission for a team, a fixed quantity of processing or a period of service? Then identify what happens when the included allowance runs out. Additional charges, reduced access and restrictions can have a greater effect on the monthly budget than the advertised starting price.
- Identify the billing unit attached to each price.
- Record the quantity included in the plan.
- Separate recurring access charges from consumption charges.
- Find the stated treatment of usage beyond the allowance.
Tokens measure text processing
A token is a unit of text representation used by a language model. A token may represent a complete word, part of a word, punctuation or another text element. Tokenisation divides text into these units before the model processes it. Consequently, 1,000 tokens do not necessarily mean 1,000 words, and the relationship between words and tokens changes with language, vocabulary, formatting and the content being processed.
Token-based pricing commonly distinguishes input tokens from output tokens. Input includes material supplied to the model, such as a prompt, instructions and relevant conversation history. Output covers the generated response. Some services also distinguish cached input or other processing categories, with separate rates. The precise definitions belong to the provider's pricing documentation.
For example, suppose an illustrative service lists an input rate of £0.20 per million tokens and an output rate of £0.80 per million tokens. A request containing 10,000 input tokens and producing 2,000 output tokens would have an estimated usage charge of £0.0036: £0.002 for input and £0.0016 for output. These figures are hypothetical and demonstrate the calculation, not a published provider rate.
Seats count people with access
A seat is a licence or allocation associated with a user account. Business software frequently uses seat-based subscriptions because organisations need to control who can access a service. The monthly charge may be calculated from the number of assigned seats, the number of active users or a minimum purchasing commitment. These definitions are not interchangeable.
For example, a plan listed at £18 per seat per month would cost £90 per month for five billable seats, before applicable taxes or additional charges. If a team has five employees but only three assigned seats, the immediate subscription calculation may use three. Whether employees can share access, transfer licences or add temporary users depends on the provider's terms.
Procurement teams should distinguish purchased seats from active users. A company paying for 20 seats may have only 14 active users during a particular month. That difference can reveal unused capacity, but removing seats may affect future access, historical records or contractual commitments.
Credits represent a provider-defined allowance
Credits are units of prepaid or allocated usage. They can be applied to image generation, video processing, voice services, model requests or a combination of features. The crucial distinction is that a credit is not a universal technical measurement. One provider's credit may represent a different quantity of work from another provider's credit, and even one provider may assign different credit deductions to different operations.
A plan advertising 500 monthly credits does not reveal the quantity of useful work those credits will purchase. Readers need the conversion rules: how many credits a request consumes, whether resolution or duration changes the deduction, and whether unused credits expire. A credit allowance may also be shared across several features, making the effective budget dependent on the user's mix of activity.
When comparing AI billing units credits, record both the allowance and the consumption schedule. If a hypothetical operation consumes 10 credits, a 500-credit allowance supports 50 such operations before considering other restrictions. If another operation consumes 25 credits, the same allowance supports only 20.
- Check whether credits reset or expire.
- Find the credit deduction for each relevant operation.
- Establish whether features share one credit balance.
- Check whether additional credits can be purchased.
Minutes measure time-based consumption
Minutes are common in audio and video services because the duration of material being processed is an important part of the workload. Speech transcription may be priced by the minute of source audio, while a voice-generation service may charge for generated audio duration. Video tools can use source duration, generated duration or a combination of duration and processing settings.
A minute-based allowance still requires a definition of what counts. Some services may round partial minutes upwards, while others may calculate usage in seconds. A transcription service might count the duration of uploaded material, whereas a real-time voice service may count active connection time or generated speech. The provider's stated measurement method determines the calculation.
For budgeting, convert the expected activity into the unit used on the pricing page. A freelancer processing eight 15-minute recordings in a month has 120 minutes of source audio to account for if the service charges by uploaded duration. A different allowance may apply if the service counts only generated speech.
Turn billing units into a monthly budget
A subscription budget should separate fixed access charges from variable consumption. Create one row for each service, then record its plan price, billing frequency, included quantity, usage unit and additional-use rate. If a service combines seats and credits, keep both measurements visible. Collapsing them into a single monthly figure makes it difficult to identify which part of the expense changes as the team grows.
For example, a small business might budget £30 for a monthly subscription, £36 for two additional seats and an estimated £12 for usage beyond an included allowance. The illustrative total is £78 before taxes and other charges. This is a planning calculation, not a statement about any particular provider's listed price.
An effective comparison also considers the billing period. An annual commitment can lower the displayed equivalent monthly price while requiring a larger payment upfront. A monthly plan can offer flexibility at a different listed rate. Record the actual payment frequency and cancellation conditions alongside the equivalent monthly amount.
- List fixed subscription charges separately.
- Calculate seat costs from the number of billable users.
- Estimate consumption in the provider's own units.
- Add a separate allowance for usage beyond the plan.
- Record taxes, currency and billing frequency where stated.
Questions readers ask
Are tokens and credits interchangeable?
No universal conversion exists. Tokens are units used in text processing, whereas credits are commercial units whose value and deductions are defined by an individual provider.
Why can two plans with the same monthly price have different usage costs?
Their included allowances, additional-use rates, feature restrictions and billing units may differ. A low subscription price can be accompanied by a smaller allowance or higher consumption charges.
What happens when a seat is removed?
The subscription may adjust at the next billing date, immediately or according to a contractual rule. Access, saved work and licence reassignment are governed by the service's account and cancellation terms.
Which unit should a freelancer track first?
Start with the unit that grows most directly with the work being sold. A text-heavy workflow may need token estimates, while transcription, voice production and video work may require duration-based records.