Pricing and comparisons

AI Power Grid publishes its exact USD price book at GET /v1/pricing. The endpoint is public, versioned, and does not require an API key. It covers text, image, video, audio, and 3D rates independently of any frontend.

curl --fail-with-body https://api.aipowergrid.io/v1/pricing

Configured price is not proof that a model is online. Check GET /v1/status/models before dispatching work.

Current same-model evidence

The price API currently publishes these narrowly matched comparisons. Evidence was reviewed on August 29, 2026 and automatically expires from the API after September 29, 2026 unless it is reviewed again.

Workload	AI Power Grid	Reference provider	Difference
`gpt-oss-120b`: 1M uncached input + 1M output tokens	$0.375	Groq: $0.75	50% lower
`z-image-turbo`: one 1-megapixel image, no optional prompt expansion	$0.003	fal: $0.005	40% lower

These are the same named models under the stated workload, not claims that every Grid model is cheaper than every hosted model. Price alone does not prove equal latency, availability, reliability, tokenizer behavior, configuration, or output quality. Compare representative outputs before choosing a provider.

⚠️

Treat comparison_evidence.status: "current" as mandatory for a public comparison claim. When the status becomes stale, the API omits comparison items while continuing to publish the Grid’s own exact rates.

Read the price book

With jq, list every configured model and its unit rates:

curl --fail-with-body https://api.aipowergrid.io/v1/pricing \
  | jq '.price_book | {version, models, aliases}'

Extract only currently valid comparison evidence:

curl --fail-with-body https://api.aipowergrid.io/v1/pricing \
  | jq '.comparison_evidence | select(.status == "current")'

Text rates are USD per one million input or output tokens. Media rates use the unit named in the response: per image, per second of video, per second of generated audio, or per completed 3D generation. The accounting ledger stores integer micro-USD, and a positive request rounds up by at most one micro-dollar.

Quote your request

The authenticated quote endpoint applies the current model price, request shape, account discount, and spendable balances without moving credits:

export AIPG_API_KEY='your-scoped-key'
 
curl --fail-with-body https://api.aipowergrid.io/v1/account/credits/quote \
  -H "Authorization: Bearer $AIPG_API_KEY" \
  -H 'Content-Type: application/json' \
  -d '{
    "model": "gpt-oss-120b",
    "modality": "text",
    "prompt_tokens": 1000,
    "max_tokens": 500
  }'

Create the key with only account.read and inference.submit. A quote is a non-mutating preflight, not a reserved allowance; the generation route still performs authoritative pricing, authorization, and balance enforcement when the request is submitted.

What the price excludes

Requests may run on remote community-operated workers. The public rate does not include a privacy or confidential-compute guarantee. Do not send credentials, personal data, or regulated content unless a separately verified deployment meets that workload’s requirements.

Worker rewards are also separate from customer pricing. Customers spend USD-denominated service credits; workers earn den accounting units and receive the active AIPG reward budget according to the supply-side settlement policy.

Use AIPG in 60 seconds Web3 Starter Projects