For prototypes & evaluation
Test
$0 / month
125,000,000 free input tokens, renewed monthly.
test_skkeys · every capability- 30 inference requests / minute
- 500,000 input tokens / minute
- No card required. Shared across your organization.
Pricing / USD
Yes. Four cents.
$0.04 per million input tokens. Output tokens are free. Start with 125 million free input tokens a month on test keys.
For production workloads
Production
$0.04 / million input tokens
prod_sk keys use purchased credits. Free test tokens stay separate.
- Buy a credit bundle of at least $10
- Auto top-up is optional
- Every capability. No output-token charge.
- Start at 1,000 inference requests / minute and 1M input tokens / minute.
Images
One image is worth
a thousand
tokens.
An image bills as a fixed number of input tokens. The detail you send sets the count. The image bills once per request. Text in the same request bills as usual.
| Capability | low · 512 px | medium · 768 px · default | high · 1024 px |
|---|---|---|---|
| yes-no, classify, classify-tree, rate | 1,000 tokens $0.04 / 1,000 images | 2,000 tokens $0.08 / 1,000 images | 4,000 tokens $0.16 / 1,000 images |
| answer | 1,500 tokens $0.06 / 1,000 images | 3,000 tokens $0.12 / 1,000 images | 6,000 tokens $0.24 / 1,000 images |
| extract, entities, verify | 2,000 tokens $0.08 / 1,000 images | 4,000 tokens $0.16 / 1,000 images | 8,000 tokens $0.32 / 1,000 images |
The input
Send the image.
- Formats
- JPEG, PNG, WebP
- Maximum size
- 5 MB
- Image storage
- None
Send bytes, a data URL, or base64. Remote image URLs aren’t supported.
The timing
Small task. Quick answer.
- Decision
- ~90 ms
answer- 0.3–0.9 s
extract- 2.5–3.5 s
Model time at medium detail.
The extraction
A few fields. Done.
5 fields per extract call
A receipt total. An ID number. A date. Image extract, entities, and verify use a generative pass for focused reading, not full-page transcription.
Capacity that grows with you
Same price.
Higher limits.
Your production tier follows your lifetime paid spend and the time since your first payment. Meet both thresholds to reach the next tier. A paid credit bundle of at least $10 unlocks production access.
| Production tier | Lifetime paid spend | Since first payment | Inference requests / minute | Input tokens / minute |
|---|---|---|---|---|
| Tier 1 | $10 minimum | Immediate | 1,000 | 1 million |
| Tier 2 | $50 minimum | 14 days | 5,000 | 1 million |
| Tier 3 | $250 minimum | 21 days | 5,000 | Unlimited |
| Tier 4 | $500 minimum | 30 days | 10,000 | Unlimited |
Each inference counts toward the rate limit. A batch of 24 texts with 3 statements counts as 72 inference requests.
Need immediate assistance with a higher tier? Reach out to our support team: support@milliseconds.ai.
Estimate production usage
What would
your workflow cost?
Set your monthly API calls and average billed input tokens per call. Start with an assumption, then use the tokens shown in a real response.
Count every call in a workflow. An invoice processed with three API calls counts as three calls here.
Estimated monthly usage
- Input tokens
- 1,000,000,000
- Per million input tokens
- $0.04
- Output tokens
- No charge
Estimated production cost
USD / month in purchased token credits
Get startedAn estimate, not an invoice. Production uses purchased credits from the first token; the free monthly test allowance is not deducted. Actual usage depends on the complete request body. See billing details.
A few billing details.
What counts as input?
The validated request body: your text plus labels, descriptions, hints, questions, types, or schema. Repeated text in separate calls is billed again. Responses include x-input-tokens so you can inspect actual usage.
What happens when free test credits run out?
Test requests pause until the next monthly renewal. Buying credits does not refill the test allowance: use a production key for paid workloads. All test keys in your organization share the same allowance.
Can production keys use the free allowance?
No. Production keys use purchased credits only. Test keys use the monthly free allowance only. Both have access to every capability.
Do I need auto top-up or a saved card for production?
No. A paid credit bundle of at least $10 unlocks production access. Auto top-up is optional: enable it if you want to replenish purchased credits automatically.
Does every capability cost the same?
Yes. All decision capabilities and the OpenAI-compatible endpoint use the same input-token price. Output tokens carry no charge. Batch requests bill the input of the items they contain.
Are demo runs included in my account usage?
Supplied examples on the demo site are free to try and can use cached responses. Custom inputs require your API key and use your account’s credits. Demos show estimated inference cost to help you plan.
How does an image get billed?
As a fixed number of input tokens, set by the detail you send: low 1,000, medium 2,000, high 4,000. The image bills once per request. Text in the same request bills as usual. Generation capabilities multiply the fixed count.
Does this include Paperwork API usage?
No. Paperwork parsing, document extraction, and redaction have their own pricing. Your account is shared, but these are separate API products.