Spark
$0
A first current you can share with a teammate.
- 1M logged tokens
- 3 keys
- 7-day traces
Pricing
Pay for logged tokens and scored rows.
Prices in USD. Annual invoices knock 15% off River and above. Write team@glimrel.store for a quote that matches your LLM volume. Model provider bills stay with you.
Most used
River
$79 / mo
Online LLM evals, caches, and Slack alerts.
- 40M tokens
- LLM-as-judge evals
- Response caches
- Slack alerts
Flood
$420 / mo
SSO, residency, and online graders at volume.
- 400M tokens
- SSO
- Data residency
- Online graders
Estuary
Talk to us
A license that stays on your floor.
- VPC
- Bring-your-own store
- Dedicated fallbacks
What a plan is buying
Gateway, traces, and evals each count inside logged tokens and scored rows. Provider inference is billed by the model host, not by Glimrel.
| Include | Spark | River | Flood | Estuary |
|---|---|---|---|---|
| Logged tokens | 1M | 40M | 400M | Custom |
| Trace retention | 7 days | 30 days | 90 days | Custom |
| LLM-as-judge evals | No | Yes | Yes | Yes |
| Multi-model routing | Yes | Yes | Yes | Yes |
| SSO | No | No | Yes | Yes |
What is always included
Email support from the founder, incident notes at team@glimrel.store, and a 30-day export if you leave. Spend caps live in the console and halt LLM traffic at the limit you set.
How a unit is counted
Logged tokens count completions that pass the gateway. Eval rows count scored spans. Cache hits still log; they do not double-bill the provider.