ChatGPT cost: subscriptions and API charges in 2026

Budgeting for ChatGPT? Separate app and team subscriptions from API charges, then cost the work needed for a catalogue-connected store assistant.

ChatGPT costs USD $20 per month on Plus, with Pro subscriptions at USD $100, USD $200 or USD $500 per month, before applicable taxes. Free access and the lower-cost Go plan are also available. A customer-facing store assistant has a separate API usage budget, plus the work of connecting and running the service.

Australian web subscriptions support local AUD pricing. The USD reference prices below aren’t AUD checkout totals. Budget the total in your billing currency, including any goods and services tax (GST) shown on the invoice.

ChatGPT costs for personal and team access

ChatGPT’s personal subscriptions pay for one person’s access, while Business charges for paid seats in a managed team workspace. OpenAI’s published subscription prices provide the following USD reference amounts. Applicable taxes are additional where charged.

AccessUSD reference priceBilling basisWhat the retailer is buying
FreeUSD $0No subscription chargePersonal access with tool limits
GoUSD $8 per monthMonthlyHigher eligible tool limits than Free
PlusUSD $20 per monthMonthlyBroader model access and expanded tools
Pro 100USD $100 per monthMonthlyPro models and higher usage allowances
Pro 200USD $200 per monthMonthlyMore included usage than Pro 100
Pro 500USD $500 per monthMonthlyHighest Pro allowance and Ultrafast access
Business StandardUSD $25 per seat per month, or USD $20 on annual billingMonthly or annual commitmentTeam workspace with central billing and admin controls
Business PremiumUSD $125 per seat per month, or USD $100 on annual billingMonthly or annual commitmentTeam workspace with higher usage limits
EnterpriseSales quoteContract termsOrganisation-wide access and additional controls

Business requires at least two paid seats, which can mix Standard and Premium. Two Standard seats therefore cost USD $50 per month on monthly billing, before applicable tax. Annual billing costs USD $480 for those two seats over a year, before applicable tax.

For the workspace’s features and data terms, read the ChatGPT Business guide. The prices above cover employee seats, with API usage billed separately.

Pro includes more than a larger message allowance. It adds Pro models and higher limits for work such as deep research. Among Pro plans, only Pro 500 includes Ultrafast for GPT-6 Astra and GPT-6.1 Sol. Model-specific allowances still apply, so a Pro subscription doesn’t mean unrestricted use of every model.

OpenAI’s multi-currency billing guidance lists AUD for Australian web subscriptions. App-store subscriptions use the relevant store’s billing. Regional prices are separate from a spot exchange-rate conversion of the USD list price. OpenAI’s subscription terms require applicable taxes to be paid.

An employee seat supplies employee access. Putting an assistant on your storefront requires its own service and integration, even when a subscription provides an API credit benefit.

Payment methods and subscription cancellation

OpenAI lists credit and debit cards for Australian web subscriptions. Purchases through the iOS or Android app use Apple App Store or Google Play billing. The monthly and annual options in the table describe billing commitments, rather than instalment finance. Your checkout shows the methods available for that purchase.

If your ChatGPT payment isn’t approved, check the card details, billing address and available funds. Complete any bank authentication prompt and ask the bank about recurring or international payment blocks. OpenAI’s declined-payment guidance also covers supported card-issuance regions. After changing a card, confirm the renewal’s payment status with the platform that billed you.

To cancel a personal web subscription, open Settings from your account menu. Select Billing, then select Cancel under Cancel plan. Cancel Apple or Google purchases in that store’s subscription settings. For self-service Business, a workspace owner uses Workspace settings, Billing, Manage plan, then Cancel subscription.

Cancel at least 24 hours before renewal to avoid the next charge. Ordinary cancellation preserves paid access until the current period ends. It doesn’t automatically refund earlier charges. Uninstalling the app doesn’t cancel the subscription, and you don’t need to delete your account to stop renewals. Follow OpenAI’s cancellation instructions and confirm that renewal is off.

API charges are a separate bill

The OpenAI application programming interface (API) charges for model usage and any separately billed tools. Its business terms make fees tax-exclusive, with applicable taxes charged separately. A token is a unit of text processed by the model. Input includes instructions, product information and conversation history. Output includes the generated response and any billed reasoning tokens.

For example, GPT-6.1 Terra’s Standard rates use these units, before applicable tax.

Token categoryUSD per million tokensWhat is charged
Ordinary inputUSD $1.00Input outside a billed cache read or write
Cached inputUSD $0.05Eligible reused context
Cache write, 30-minute retentionUSD $1.25Context written for reuse
Cache write, 12-hour retentionUSD $2.50Context written with longer retention
OutputUSD $5.00Billed output tokens

These rates assume requests below the model’s long-context threshold, without Fast mode or regional-processing premiums. A different model changes the calculation. Cache-write charges replace the ordinary input rate for those tokens. They aren’t an extra charge on top of ordinary input.

OpenAI’s API tool pricing lists file search at USD $2.50 per 1,000 calls. Storage costs USD $0.10 per GB per day after the first free GB. Retrieved content also contributes to model input. A store using its own search service must budget that provider’s charges instead of assuming hosted retrieval is included.

ChatGPT and the API retain separate billing systems. Eligible paid Pro and Business subscriptions can nevertheless claim API credits. The subscription must be at least seven days old, and its owner must link an API organisation they also own.

Pro’s monthly API credit amounts are USD $100, USD $200 or USD $500, matching the tier. Business provides USD $20 per Standard seat or USD $100 per Premium seat, subject to a workspace cap. Credits have billing-cycle expiry rules. They offset eligible API usage after activation, without turning the app subscription into an unlimited storefront service.

Purchased prepaid credits fund API usage in advance and expire after one year. The minimum purchase is USD $5. A positive balance doesn’t remove model access requirements, request-rate limits or approved usage limits. An automatic-recharge limit controls credit purchases, and isn’t an instantaneous spending cutoff.

Estimate a catalogue conversation workload

A hypothetical store with 5,000 conversations a month and four model responses per conversation makes 20,000 model calls. Assume each call averages 2,000 input tokens and 300 billed output tokens. The input allowance includes product context, instructions and earlier messages.

That produces 40 million input tokens and six million output tokens. For this example, split input into 30 million ordinary tokens, eight million 30-minute cache-write tokens and two million cache-read tokens. This is an illustrative billing mix, not measured store traffic or a promised cache hit rate.

Use GPT-6.1 Terra at the Standard rates above. Assume one hosted file-search call per response and two GB of vector-store contents held throughout a 30-day period. One of those GB is free. All amounts in this example exclude applicable tax.

Cost itemHypothetical monthly calculationUSD cost
Ordinary input30 million tokens at USD $1.00 per millionUSD $30.00
Cache writesEight million tokens at USD $1.25 per millionUSD $10.00
Cache readsTwo million tokens at USD $0.05 per millionUSD $0.10
OutputSix million tokens at USD $5.00 per millionUSD $30.00
File-search calls20,000 calls at USD $2.50 per 1,000USD $50.00
File-search storageOne billed GB at USD $0.10 per day for 30 daysUSD $3.00
Usage totalModel tokens plus hosted retrievalUSD $123.10

The estimate is about USD $0.02462 per conversation before tax, hosting and support. It counts conversations, including those that end without an order. It is not a cost per sale.

One extra response in every conversation adds 5,000 model calls and 5,000 file searches. With 2,000 ordinary input tokens and 300 output tokens per added call, that adds USD $30 before tax. The total becomes USD $153.10, assuming storage stays unchanged and the added input receives no cache writes or reads.

Product context can grow without conversation counts changing. Adding 1,000 ordinary input tokens to each of the original 20,000 calls adds USD $20 before tax. Cached context has a different rate, so count the billing categories separately. Deduct any activated, eligible API credits only after calculating gross usage.

Include the work beyond model usage

Running a catalogue-connected assistant also requires current product data and a maintained connection to your store. Catalogue sync must handle stock and price changes. Policy documents need a process for keeping delivery and returns answers accurate.

Buying in chat adds basket and payment connections, with order confirmation and recovery when a step fails. After-sale help needs authenticated access to the right orders and a path to staff when automation can’t resolve a request. Those connections need engineering and ongoing support.

Hosting and monitoring remain separate budget items. Include conversation storage, error alerts and checks that catch incorrect product or policy answers. Allow staff time for catalogue changes and repairs when a commerce connection changes. A low token bill alone doesn’t price those responsibilities.

For other vendors’ billing models, use the customer service software pricing directory.

A customer browsing products inside ChatGPT is using a different channel from an assistant embedded on your own site. The ChatGPT subscription table prices user access to ChatGPT. It doesn’t quote the work of running your store’s service or establish merchant participation terms.

Compare a build with a ready-to-use shopping agent

shoppa is the stronger choice for retailers who want product discovery and buying in chat on their own site without building the shopping service. ChatGPT suits staff who want general research, writing and analysis. An API build is a separate engineering choice for a team that wants to own that implementation.

With shoppa, customers refine product choices through a conversation that keeps their earlier requirements. Answers use your catalogue’s prices and stock, alongside your own policy documents. The agent runs on your site with your branding, so the product conversation stays with your store.

Customers can compare products, add them to their basket and pay in the chat. They can then check orders or start returns, with human handover when they need a person. Those connected steps are part of shoppa’s shopping service. In a custom build, your team must implement and maintain the equivalent flow.

Compare both routes using the same conversation volume and catalogue freshness requirements. Include checkout, order support and staff handover in each scope. Otherwise, a model-only estimate will look cheaper because it leaves part of the service unpriced.

shoppa has per-session pricing for predictable session costs and per-token pricing tied to what the agent uses. For a retail service budget, get a quote covering your expected traffic and store connections. A ChatGPT seat price alone cannot provide that comparison.

Choose the access that matches the job

Choose a personal ChatGPT plan for your own research or drafting, with the tool limits your work needs. Business fits employee access when you need a managed workspace and central billing. Enterprise requires a sales quote for the access and controls in its contract.

For a customer-facing service, estimate API usage from the calls and context the service will actually send. Add retrieval and operating costs, then account for any eligible credits. Ask an implementation provider to price the catalogue and commerce connections alongside continuing support.

For a ready-to-use shopping agent on your own storefront, price shoppa against that complete service scope. Keep employee ChatGPT subscriptions as a separate budget line so staff access and shopper assistance each have a clear cost.

Give your store its own shopping agent

Customers ask in their own words. Your agent finds, compares and checks out from your catalogue, right on your site.

See it in actionFree trial · Simple setup