Subscription vs. Per-Token AI Pricing

Per-token APIs bill for tokens in individual requests, while Borg provides access to models through a subscription. To compare them, estimate your expected API spend and compare it with the current subscription plan.

Two ways to pay for AI models

Each model maker sets API rates, which can differ between input and output tokens. Your total changes with the amount and type of model use. Borg provides access to the models in its directory through a subscription.

Prices on the model pricing page are vendor API list prices, not extra per-request charges from Borg. Check the current Borg plan for subscription details.

Do not treat a model's vendor list rate as Borg's price.

An API bill depends on request volume and token use. A subscription has a plan price, so compare the plan with your expected API spend.

How to read vendor prices

Vendor prices are often listed separately for input and output, as a rate per million tokens. Input tokens are the text or other supported content sent to a model. Output tokens are the response it generates.

For an API estimate, multiply the relevant token volumes by the listed rates, then add input and output costs. Rates vary across models and vendors. The directory shows cited rates where available, alongside a source.

A price marked “Not published” means the catalog has no cited vendor list price. It does not mean the API is free. Check the vendor before using a price in a budget.

When an API may cost less

A per-token API may cost less when low usage keeps metered charges below the subscription price. It may also fit when you need to pay a vendor directly for a particular API workflow. The result depends on model, request mix, input and output volume, and rates.

Estimate your input and output use separately. If a source price is missing, do not assume zero cost. Verify the vendor's current rate before deciding.

When a subscription may cost less

A Borg subscription may cost less when its price is below your expected vendor API spend. It may also suit people who want access to models in the directory without tracking a separate per-request bill for each vendor API. Compare the plan with your own expected use.

Borg Fusion can choose a model for each request, or you can select a model lane directly. See the Borg Fusion page and model pages. Your model and routing choices affect an API estimate, so include the models you expect to use.

The right option depends on your usage and model mix.

Compare your own costs

Use the Cost Savings Calculator to compare expected usage. Treat the result as an estimate, and check current plan details and vendor rates before deciding. Vendor list prices can change, and a catalog entry may not include a cited price for every model.

Estimate your metered API bill, compare it with the current subscription price, and account for the models you need. Neither a subscription nor per-token pricing is automatically cheaper for everyone.

FAQ

How does per-token API pricing work?

A vendor API bills based on tokens processed by individual requests, using rates set by that vendor.

Does Borg charge vendor API rates for each request?

No. Borg provides access to the models in its directory through a subscription; the listed vendor rates are for comparison.

Which option costs less?

It depends on usage and the subscription plan. Compare your expected API usage with Borg's current plan and use the Cost Savings Calculator.