Pricing

Celeris uses prepaid, per-token billing. Buy credits under Settings → Billing in the Console; completed requests draw from the workspace's balance according to token usage. There are no subscriptions, seat charges, or minimum commitments.

Rates

Each model is priced separately, per million tokens:

Model	Prompt tokens	Cached prompt tokens	Completion tokens
celeris-1	US$0.20 per million	US$0.02 per million	US$0.70 per million
celeris-1-magnus	US$0.20 per million	US$0.02 per million	US$0.70 per million

Estimating costs

Completed responses report exact token counts in their usage block:

"usage": {"prompt_tokens": 27, "completion_tokens": 2, "total_tokens": 29}

At celeris-1's rates, cost per call is prompt_tokens × $0.20/M + completion_tokens × $0.70/M. The following examples show approximate costs for common agentic workloads:

Workload	Prompt	Completion	Cost per call	Calls per US$1
Classification / routing	~30 tokens	~4 tokens	~$0.0000088	~110,000
Judge / rerank scoring	~120 tokens	~3 tokens	~$0.0000261	~38,000
Structured extraction	~200 tokens	~48 tokens	~$0.0000736	~14,000
Query rewriting	~60 tokens	~24 tokens	~$0.0000288	~35,000

Two request settings have the greatest effect on spend:

• Set max_tokens deliberately. Completion tokens cost 3.5× as much as prompt tokens, and a limit bounds the cost of unexpected output. Any positive integer that fits the context window is accepted.
• Keep prompts concise. Repeated instruction tokens are billed on every call.

Images are billed as prompt tokens at the same rate, with no separate per-image charge. An image's token cost tracks the resolution the model processes it at rather than its file size, so recompressing an upload to fewer bytes cuts latency but not tokens. Read the exact count from the response's usage block.

An image extraction, worked through

The Console's Read an image example is a typical shape for image work: a short instruction carrying an output schema, one receipt image, and a small JSON reply. Its parts land roughly here:

Part	Direction	Tokens
Instruction and output schema	Prompt	~140
One receipt image	Prompt	~275
JSON reply	Completion	~48

The reply is sized like the structured extraction row above: a handful of short JSON fields. That comes to ~415 prompt tokens and ~48 completion tokens, so 415 × $0.20/M + 48 × $0.70/M ≈ $0.0001166 per call, or roughly 8,600 calls per US$1.

The image is about two thirds of the prompt here, so on an image workload what moves the bill is how many images each call carries, not how long your instructions are. The reply is bounded by max_tokens either way: at the 2,048 default, the worst case for the same call is ~$0.0015.

These are one example's figures rather than a rate card. Your image, your prompt, and how much the model writes back all move them. Measure your own first: Token usage and cost shows how to read an image's real cost out of a response.

What is — and isn't — charged

• Served requests are charged from their exact usage token counts, the same token volumes the usage dashboard aggregates.
• Rejected requests are not charged. This includes throttled (429), unauthorized (401), and out-of-credit (402) requests. The exception is a JSON mode request that fails its contract: that 400 carries usage, and the generations it reports are charged.
• Server-side failures are free. You are never charged for a 502 or 503.
• A client disconnect does not guarantee cancellation. If processing continues after a disconnect, the request may still be charged for work performed.

Buying credits

Buy credits with a card (Stripe) under Settings → Billing, in preset amounts of $5 / $10 / $50 / $100 or a custom amount between $5 and $1,000 per purchase. The card must be issued in a country where Celeris is available (currently the United States); see Availability if you're somewhere else. Receipts are emailed automatically, and invoice documents remain available on the same billing page.

When your balance reaches zero, API calls return 402 insufficient_quota until you add credit. See Going to production for production monitoring guidance. See Auto Top-Up for automatic purchases and the balance-email schedule.

Credit expiry

Credit granted to your workspace expires 30 days after it is granted.

Usage draws from the credit that expires first. Expired credit leaves your balance and cannot be restored; from a zero balance, API calls return 402 insufficient_quota until you add credit.

Settings → Billing shows the next date on which some of your credit expires.

We also email you about it: once two weeks ahead if you have not spent any of the credit yet, once three business days before the date, and once when it expires. Turn these off under Credit balance and expiry in Settings → Notifications.

For reserved capacity, custom limits, or invoicing beyond self-service top-ups, contact us.