Pricing

Pay per token from a prepaid balance, or buy a budget pack once. Prices come from the active price version and apply from the moment they are published.

Budget packs

Up to 90% off · selected models

10 models

$1.99

one-time

Worth up to ≈ $20 at official prices

See all 10 models
  • 10 models
  • Budget = price paid, never expires
  • Pay with QRIS or USDT/USDC, no auto-renewal
  • One API for OpenAI and Anthropic formats
  • Billed per token, cache counted
Up to 93% off · selected models

29 models

$9.99

one-time

Worth up to ≈ $161 at official prices

See all 29 models
  • 29 models
  • Budget = price paid, never expires
  • Pay with QRIS or USDT/USDC, no auto-renewal
  • One API for OpenAI and Anthropic formats
  • Billed per token, cache counted
Up to 93% off · selected models

38 models

$99

one-time

Worth up to ≈ $1,597 at official prices

See all 38 models
  • 38 models
  • Budget = price paid, never expires
  • Pay with QRIS or USDT/USDC, no auto-renewal
  • One API for OpenAI and Anthropic formats
  • Billed per token, cache counted

“Worth” is the same tokens at official prices, on the models with the largest discount.

Pay as you go

Top up the balance and pay per token. The balance does not expire.

  • — Four rates per 1M tokens: input, cache read, cache write, output
  • — Each request is rounded up to the next nano-dollar, at least $0.00002

How a pack works

  1. A budget is not cash

    It pays for requests at Cent AI prices on the models in the pack. It cannot be withdrawn or refunded.

  2. When the budget runs out

    Requests continue from your PAYG balance. With no balance left, requests stop with 402.

  3. Unused budget stays

    The budget does not expire. Buy the same pack again whenever you want more.

Allowance per model

EstimateAssumes 99% input and 1% output tokens; the cached view reads 90% of input from cache. Billing always follows real usage. Worth: those tokens at each model’s official price.

Printing the allowance…

Image models

Priced per image or per token. Not part of budget packs.

Image models
ModelDoesPricevs official
Hunyuan Image Alpha
Checking status
New images from text
$0.15Official price $1.00
85% off
Hunyuan Image Alpha Edit
Checking status
Edits one image you upload
$0.15Official price $1.00
85% off
MiniMaxMiniMax H3 Image
Checking status
New images from text
Text · Per 1M input tokens$0.345Official price $2.30Image · Per 1M input tokens$0.225Official price $1.50Image · Per 1M output tokens$0.57Official price $3.80≈ Per image $0.003174
85% off
Alibaba QwenQwen Image 3.0 Pro
Checking status
New images from text
$0.021Official price $0.04
47% off
Wan 2.7 Image
Checking status
New images from text
$0.02Official price $0.03
33% off
Wan 2.7 Image Pro
Checking status
New images from text
$0.049Official price $0.075
34% off
  • New images from text

    Per imageOfficial price $1.00$0.15

    Status: Checking status
  • Edits one image you upload

    Per imageOfficial price $1.00$0.15

    Status: Checking status
  • New images from text

    Text · Per 1M input tokensOfficial price $2.30$0.345
    Image · Per 1M input tokensOfficial price $1.50$0.225
    Image · Per 1M output tokensOfficial price $3.80$0.57
    ≈ Per image$0.003174

    Status: Checking status
  • New images from text

    Per imageOfficial price $0.04$0.021

    Status: Checking status
  • New images from text

    Per imageOfficial price $0.03$0.02

    Status: Checking status
  • New images from text

    Per imageOfficial price $0.075$0.049

    Status: Checking status

What can surprise you on the bill

  • Reasoning tokens count as output

    Models that think before answering bill those tokens at the output rate, as the official price does.

  • A cancelled stream is charged

    The input and the tokens the model sent before you disconnected are charged.

  • Cache is never guaranteed

    Cache rates apply only when the upstream reports cached tokens; otherwise the whole input is billed at the input rate.

  • Failed requests cost $0

    A request that fails before the model receives it is not charged.

  • A minimum per request

    Each successful request costs at least the model's minimum charge, $0.00002 unless the model says otherwise.

  • The rupiah rate is locked at top-up

    A QRIS top-up converts rupiah to USD at the rate shown when you create the order.

Pricing questions

Can a price change after I pay?

Prices are versioned. A request uses the version active when it arrives; a new version, up or down, applies from the moment it is published on the changelog.

Which is used first, the pack or the balance?

Promo credit first, then the pack budget for the models it includes, then the PAYG balance.

Can I hold more than one pack?

Yes. Each pack keeps its own budget, and the lowest pack that includes the model is used first.

Is there a fee on top-ups?

Cent AI pays the QRIS fee. For USDT/USDC, the network fee is paid by the sender in their wallet.