Building at scale?
Discuss model access, capacity and billing requirements.
Or write to hello@getaight.ai. Include the company, the expected monthly volume and the models in use.
- Provider list prices
- Higher rate limits
- Some models run in India
- models in one API
- 139
- model makers
- 15
- days' notice before a price rise
- 7
- days' notice before a model is removed
- 14
What can be arranged
Model access
Models from the catalogue, the regions they run in, and which keys may call them.
Capacity
Higher limits for production traffic, planned around the expected volume.
Billing
Prepaid balance in US$, GST tax invoices for India accounts, and one balance shared by every key.
Prompts pass through. They are not kept.
Aight records the facts needed to bill a request, never its content.
Prompts and outputs are not stored
They pass through Aight's servers only for the time it takes to get a reply and are not written to disk.
Never used for training
Microsoft Azure, Google Vertex AI and Amazon Bedrock each state in their business terms that they do not train on customer data.
The region is known before the request
Every model is tagged India, United States or Global on Models & pricing. Aight's own servers run in India.
Keys stored hashed
A key is shown once and cannot be read back. A leaked key can be frozen in the console or by support.
Logging only on request
Debug logging is off by default. When turned on, prompts and outputs are kept for 7 days.
Metadata, for the bill
Time, key, model, token counts, cost and status. Kept one year after the last activity.
Billing a finance team can file.
- Invoices
- A GST tax invoice in ₹ for every recharge from an India account, with the GSTIN on the account. Accounts outside India get a US$ invoice. Every invoice is in the console and sent by email.
- Balance
- Prepaid and held in US$, shared by every key on the account. It does not expire.
- Exchange rate
- India recharges convert at the RBI reference rate plus 2%, shown before payment and printed on the invoice. Terms
- Refunds
- Unspent paid balance goes back to the original payment method, in the currency paid, at the original rate. Extra is not refunded. Refunds
- Changes
- Price rises are emailed at least 7 days ahead. Retired models get 14 days' notice where the provider allows, with a named replacement.
How it starts.
Write to Aight
The company, the expected monthly volume and the models in use.
Verify the email
The account opens once the email is verified. Anonymous use and resale of access are not allowed.
Keys and recharge
Create keys with caps and allowed models, then recharge.
Questions procurement asks.
Is the API compatible with existing code?
Yes. The API follows the OpenAI format. Official OpenAI SDKs work with the Aight base URL and an Aight key.
Can spending be limited per team or project?
Give each team its own key. A key can have a monthly cap, an expiry date and a list of allowed models. An empty balance stops every key.
Where are requests processed?
In the region shown for each model on Models & pricing: India (the provider's South Asia data zone), United States, or Global.
What happens when a model is retired?
At least 14 days' notice is given where the provider allows it, and a replacement is named. If none suits, unspent paid balance can be refunded.
What if a key leaks?
Freeze or delete it in the console at once, or write to support with "key leaked" as the first words.
Can the account be closed?
At any time, by writing to support. Unspent paid balance is refunded.
Plan the rollout with Aight.
hello@getaight.ai