Pricing
Pay per token, from prepaid credits.
Get started
$20
- Your agent signs up in one command
- Top-ups from $20, approved by your human
- Every call is paid from prepaid credits
- US region (Canada coming soon)
Prepaid credits
From $0.22/ 1M input tokens
How credits work
- 1
Your agent asks
It requests a top-up and gets an approval link.
- 2
Your human approves
They open the link and pay by card.
- 3
Usage draws it down
At each model's per-token price. Top up again any time.
Per model, per token
Input, cached input and output, priced for each model.
Charges need approval
When credits run out, requests stop with a 402 and nothing is invoiced. Nothing is charged without your human's approval.
Credits don't expire
The balance waits until you need it.
Every live model
Priced per model
Each model has its own price in USD per 1M tokens: input, cached input and output. Batches cost 50% of the list price. Your agent sees every live model's price with npx -y scaledthought@0.3 models --live.
Fine-tuning (LoRA) is billed per 1M training tokens (dataset tokens × epochs), at a rate set by the base model's size. A new job returns its estimated cost and holds it from credits; the difference is settled when it finishes, and a failed or cancelled job is refunded in full. A fine-tuned model costs the same per token as its base.
| Model | Context | Input | Cached input | Output |
|---|---|---|---|---|
| MiMo-V2.6-FlashXiaomimimo-v2-6-flash | 1M | $0.22 | $0.0044 | $0.44 |
| DeepSeek-V4.1-FlashDeepSeekdeepseek-v4-1-flash | 1M | $0.22 | $0.007 | $0.66 |
USD per 1M tokens, US region, from the US region's API. Refreshed at most every 5 minutes; last read . Cached input is billed at the input price where a model has no cached rate. The same list from the CLI: npx -y scaledthought@0.3 models --live.
Rate limits
Limits per key
New account
60
requests / min
200,000 tokens / min
Until the first credit purchase: up to 5 keys
Prepaid credits
3,000
requests / min
5,000,000 tokens / min
Credits your human buys, $20 to $2,000 per top-up, $2,000 per 24 h
Scale
Custom
requests / min
Custom tokens / min
Talk to us
Questions
Billing questions
- Do I need a card to sign up?
- Not to create the account: your agent does that in one command. Every call is paid from prepaid credits, so your human pays by card when they approve the first top-up, before the first call.
- How does billing work?
- Prepaid credits, $20 to $2,000 per top-up, at most $2,000 per 24 hours and 2 open approval links at a time. Your agent runs billing topup, which returns an approval link for your human. They pay by card and the credits land once the payment is confirmed. Usage is taken from the balance at each model's per-token price.
- What happens when credits run out?
- Requests return 402 insufficient_credits with the balance and the amount needed, and nothing more is spent. Your agent asks your human for another top-up. There is no bill at the end of the month.
- How much can I buy at once?
- $20 to $2,000 per top-up, at most $2,000 per 24 hours.
- What is auto-reload?
- Optional. When the balance drops below a threshold, we charge the saved card a fixed amount, up to a monthly limit. Your human approves all three numbers in the same link. Turning it off or lowering it needs no approval; if a charge is declined, auto-reload switches itself off.
- What needs credits?
- Everything: every API call, batch and fine-tuning job is paid from prepaid credits. After the first purchase you also get higher rate limits and more than 5 keys.
- Do credits expire?
- No. Unused credits stay on the account.
- What are cached input tokens?
- Repeated prompt prefixes (system prompts, long documents) are cached automatically and billed at the model's lower cached-input rate.
- How are prices set?
- Per model and per region, in USD per 1M tokens. The table above lists every model live in the US region with the price it charges now, read from the US region's API.