COOLJEV / PRACTICAL GUIDE
Jev pricing: calculate API cost per workflow
Copy this cost calculation
records = 10_000
input_tokens_per_record = 1_000 # planning assumption, not measured usage
usd_per_million = 0.042
print(records * input_tokens_per_record / 1_000_000 * usd_per_million)
This hypothetical workload costs $0.42 in provider input charges at that rate. It is arithmetic, not a live billing result. Use the cost calculator to change your assumptions.
Which price is this?
The official model reference lists Jev 1.13 at $0.042 per million input tokens, with output tokens free. Confirm the current provider account terms before purchasing or budgeting. Gateway prices and account arrangements may differ.
| Budget item | How to count it |
|---|---|
| Input | Total request input, including state and questions |
| Repeated requests | Add each newly billed call; do not count only unique records |
| Preprocessing | Separately price transcription, OCR or other services |
| Human review | Review volume multiplied by time per item |
| Integration | Logging, testing, storage and maintenance |
Character counts are estimates, not tokenizer measurements. After permitted live runs, collect actual usage.input_tokens and compare it with your estimate. A long shared policy or conversation can dominate input size even when the decision is one label.
CoolJev credits are a separate unit
The hosted workspaces use one credit per submitted item; several questions for that item share a request. Guest and signed-in quotas are displayed in the workspace. These beta credits are not TypeSafe currency, a paid subscription or a promise of free official API access. Downloaded live examples use your own provider key and provider billing.
Decide whether lower token cost matters
Measure the full workflow on the same labeled records. A cheaper request can still cost more if it creates many false escalations or requires more review. Record errors, retry counts and unresolved records alongside cost. Compare a rules baseline before spending time on a model integration. No latency or accuracy benchmark is claimed here.
Independently written by CoolJev. Sources checked and updated: 2026-09-24