Multi-model AI · one prepaid balance
Intelligence,
on your terms.
Chat, build, and manage your AI usage from one workspace. Choose your model, control your budget, and pay for what you use.
- Catalogue
- 6 models
- Billing unit
- $0.000001
- Commitment
- None · prepaid
Add credits
Buy prepaid service credits by card. One credit is one US dollar of usage. Credits appear after the payment is confirmed server-side.
Choose a model
Pick from the catalogue per conversation or per API call. Every model lists its price per million tokens.
Track usage
Each request records tokens, cost, latency and status. Filter by date, break down by model, export CSV.
How a request flows
From a question to an itemised cost.
01
Ask in the workspace
One composer for every model. Write here and continue in your workspace — conversations are saved to your account and searchable by title.
02
Choose the model for the job
Switch per conversation. Prices are shown per million tokens before you send, and the context limit is checked for you.
Prices per 1M tokens, read from the catalogue database.
03
Know the cost to the millionth
Tokens in, tokens out, the price version used and the exact amount debited from your prepaid balance. Try the formula with your own numbers.
- Model
- Claude Opus 5
- Rate (in / out per 1M)
- $5.00 / $25.00
- Reserved while it runs
- ≥ $0.016000
Same arithmetic as the server: integer micro-dollars, rounded up once per request. The real reservation uses a deliberately higher input estimate.
Model catalogue
The models on this deployment, exactly as configured.
This list is read from the catalogue database. A model is offered only after a live test request has succeeded with the operator's provider key — until then it is shown with the reason it is unavailable.
| Model | Upstream ID | Context | Input / 1M | Output / 1M | Status |
|---|---|---|---|---|---|
Claude Opus 5 Deep reasoning and long, complex work. Spends hidden reasoning tokens, billed as output. | Anthropic · claude-opus-5 | 1M | $5.00cached $0.50 · write $6.25 | $25.00 | Available |
Claude Sonnet 5 Balanced speed and quality for everyday work. Spends hidden reasoning tokens, billed as output. | Anthropic · claude-sonnet-5 | 1M | $2.00cached $0.20 · write $2.50 | $10.00 | Available |
Claude Haiku 4.5 Fast and inexpensive for short tasks. | Anthropic · claude-haiku-4-5 | 200K | $1.00cached $0.10 · write $1.25 | $5.00 | Available |
GPT-5.6 Sol High-capability general model. Spends hidden reasoning tokens, billed as output. Limited here to 272K input tokens. | OpenAI · gpt-5.6-sol | 272K | $4.00cached $0.40 · write $5.00 | $20.00 | Available |
GPT-5.6 Terra Mid-tier general model. Spends hidden reasoning tokens, billed as output. Limited here to 272K input tokens. | OpenAI · gpt-5.6-terra | 272K | $2.00cached $0.20 · write $2.50 | $12.00 | Available |
GPT-5.6 Luna Low-cost model for light tasks. Spends hidden reasoning tokens, billed as output. Limited here to 272K input tokens. | OpenAI · gpt-5.6-luna | 272K | $0.20cached $0.02 · write $0.25 | $1.20 | Available |
Text chat with streaming. Images, files and tool calling are not offered. Cached-input and cache-write rates apply only when the provider reports those tokens for a request.
Pricing
Usage-based. Itemised to the millionth of a dollar.
There is no subscription and no bundled allowance. You hold a prepaid balance of service credits; each request is priced from the tokens the provider reports, at the per-model rates in the catalogue.
Credit packs · 1 credit = US$1 of usage. Card payments are running in the payment provider’s sandbox on this deployment.
- 01
Reserve
Before a request is sent, a conservative maximum cost is reserved from your balance, so concurrent requests can never spend the same credit twice.
- 02
Settle
When the response ends, the cost is calculated from the provider’s reported usage at the price version captured when the request started, and rounded up once to $0.000001.
- 03
Release
Whatever part of the reservation was not used returns to your available balance immediately.
- 04
Reconcile
If a response is stopped or interrupted before the provider reports usage, the reservation is held as “pending” and settled from a labelled estimate — never silently dropped, never invented as exact.
Credits are prepaid service credits for EdifacePay usage. They are not a bank deposit, are not withdrawable, and earn no interest.
Developers
One key, the same balance, a documented API subset.
GET /v1/models and POST /v1/chat/completions with streaming text, in the OpenAI request shape. It is a subset: unsupported parameters are rejected with a clear error rather than ignored.
- · Keys are shown once and stored only as a keyed hash
- · Per-key monthly and lifetime spending limits, expiry and revocation
- · API and workspace requests share one reservation and ledger path
curl https://www.edifacepay.org/v1/chat/completions \
-H "Authorization: Bearer $EDIFACEPAY_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "claude-opus-5",
"stream": true,
"stream_options": { "include_usage": true },
"messages": [{ "role": "user", "content": "Hello" }]
}'- Available
- —
- Reserved
- —
- Monthly limit
- —
Reserved credit belongs to requests in flight. It returns to “available” the moment they settle for less. Create an account to see your own figures here.
Budget control
Spend limits that are enforced before the request, not after the bill.
- Daily and monthly limits on your account. A request that would cross a limit is refused up front.
- Per-key limits for every API key, so an integration can only spend what you allow it.
- Output fitted to your balance. In the workspace, a reply is capped to what your credits cover — and you are told when that happens.
- Full visibility. Requests, tokens, cost and latency by model and by day, with CSV export.
Token
$EPAY
The EdifacePay community token. It is separate from the product: credits are bought by card, and $EPAY is not a payment method, not required to use the service, and not an investment offer.
Follow @edifacepay for the announcement- Ticker
- $EPAY
- Network
- Robinhood Chain
- Contract
0x4d76f021c7cc83892f7d69cdb913ab2264c9fe8d
The address above, shown on edifacepay.org, is the only official $EPAY contract. Check it character by character before any transaction.
Questions
Straight answers.
What exactly am I buying?+
Prepaid service credits for EdifacePay. One credit pays for one US dollar of usage at the listed per-model rates. Credits are not a bank balance, cannot be withdrawn, and have no value outside this service.
When do purchased credits appear?+
When the payment provider confirms the payment to our server through a signed webhook. The browser redirect after checkout never adds credits, so a closed tab or a replayed link cannot change your balance.
What happens if I stop a response halfway?+
We cancel the upstream request. Providers still bill what was generated before the stop, and they do not always report that figure. The reservation is kept as “pending” and then settled from an estimate that is labelled as an estimate on your usage record; the rest is released.
Does regenerating a reply cost extra?+
Yes. Regenerating sends the conversation to the model again, and the new request is billed like any other. The workspace says so before you confirm.
Where are my conversations stored, and who processes them?+
Conversations are stored in this service’s database so you can return to them, and you can delete any conversation — that removes its messages from our database. To answer you, the conversation is sent to the provider of the model you chose (for example Anthropic or OpenAI) and is processed under that provider’s own terms and retention policy. We cannot delete data on a provider’s side and do not claim to.
Is there an unlimited plan?+
No. Everything is usage-based against your prepaid balance.
Is EdifacePay affiliated with the model providers?+
No. EdifacePay is an independent service that calls provider APIs with its own keys. Model names belong to their respective owners.