▶ modelweave.io / pricing

You pay per 1K tokens replayed. Never per seat.

The wedge is honest: the cost of running ModelWeave should be a small, predictable fraction of the savings it surfaces. If we do not save you the replay fee, the bill is zero.

tier_00

Solo

$0/ forever

Kick the tires. Replay your last 30 days against three saved rules.

  • Replay window: 30 days
  • Saved rules: 3
  • Bayesian Thompson posterior on every counterfactual
  • SSE stream: 60 lines/min
  • Community Discord
open /replay
tier_01

Pro

$0.0004/ 1K tokens replayed

For the SRE who replaced a $14K/mo Anthropic invoice with a $9K one and wants the audit trail to prove it.

  • Unlimited replay window
  • Unlimited saved rules
  • Cmd+K rule library + recipe sharing
  • PagerDuty + Datadog + Honeycomb webhooks
  • Edge KV at 5ms p99 (K3.2)
  • Email + Slack support
start replay billing
tier_02

Firm

invoice/ quarterly

SAML, audit logs, on-call SLA. For the platform team that signs MSAs.

  • Everything in Pro
  • SAML SSO + SCIM
  • Audit log export (S3/CloudWatch)
  • 99.95% SLA, 1hr P1 response
  • Dedicated Slack Connect
  • Private VPC deployment option
book a call

What you are actually paying for

the painthe resolution
the OpenAI invoice ticked over $50K last monthreplay 30 days against a cheaper-arm rule → see the dollar delta
the new model claim is "30% better" with no mathThompson posterior 95% CI — claim is now a number with a CI
the routing logic is buried in 4,200 LOC of middlewarerules live as toggle chips; toggle = view-transition rebind
Datadog shows latency, not moneythe log rail prints $ saved per decision, ANSI by provider