VoidAI Credits, Pricing, and Cache Billing

VoidAI charges credits according to the model and operation. Browse current models and prices, or inspect the model API. Model rates and eligibility can change; the examples below explain the calculation rather than quote current prices.

Token-based pricing

Input and output can have different rates. Cache reads and cache writes are separate portions of the total input:

uncached input = total input - cache-read input - cache-write input

raw cost = base_cost
         + uncached input × input_multiplier
         + cache-read input × input_multiplier × cache_read_multiplier
         + cache-write input × input_multiplier × cache_write_multiplier
         + output tokens × output_multiplier

credits = round(raw cost / applicable discount multiplier)

If the model has no cache-write multiplier, the calculation uses 1. If no discount greater than 1 applies, the cost is not divided. Do not add cache tokens to the total input a second time. For an illustrative model with input rate 1, output rate 3, cache-read multiplier 0.1, cache-write multiplier 1.25, and base cost 0:

Input or output	Tokens	Credits
Uncached input	600	600
Cache-read input	300	30
Cache-write input	100	125
Output	200	600
Total before a discount		1,355

With an applicable 2× discount, this example rounds to 678 credits.

Context and service tiers

Where a model exposes long-context rates, those rates apply when pricing input is strictly greater than long_context_threshold. A supported priority service tier can also increase rates. The catalog does not expose every service-tier adjustment. Usage fields differ between API formats. In native Messages responses, input_tokens can exclude the separate cache_read_input_tokens and cache_creation_input_tokens. Billing can use a local input estimate when it exceeds reported input usage, so raw upstream counters alone may not reproduce the final charge. Use request usage to inspect credits_used.

Fixed-cost operations and images

Models with cost_type: "fixed" use a base cost rather than the token formula, subject to the operation’s billing rules. Image generation/editing is charged per image: non-streaming requests use the number of returned images; completed streaming requests use the requested count. Do not assume one request always costs one image’s price. See each image endpoint for access requirements. A model’s appearance in the public catalog is not a guarantee of eligibility.

Balance and discounts

View your balance and plan in the dashboard, or call GET /v1/credits.
Read active model-specific discounts with My Discounts.
Use request usage for the recorded spend on a completed request.

Rate limits and account entitlements vary. Handle HTTP 429 using the response’s retry guidance rather than assuming a universal requests-per-minute allowance.

Getting Started

Guides

VoidAI Credits, Pricing, and Cache Billing

Token-based pricing

Context and service tiers

Fixed-cost operations and images

Balance and discounts

​Token-based pricing

​Context and service tiers

​Fixed-cost operations and images

​Balance and discounts

Token-based pricing

Context and service tiers

Fixed-cost operations and images

Balance and discounts