Platform · Foundry capability
Model routing
No single model is best at everything. Foundry routes every task to the right model, frontier to local, balanced for intelligence, accuracy, cost and privacy.
Frontier model partners
Every task, the right model.
The router weighs each task against policy and fit, then sends it where it runs best, with citations carried back.
Work
- Draft a clause
- Reconcile 4,000 lines
- Classify an invoice
- Summarise a tender
Edrak Foundry
the router · every task balanced across four axes
- Intelligence
- Accuracy
- Cost
- Privacy
Policy narrows the field first: residency and data classification decide which models a task may even reach.
Models
- Frontier models
- Fast & efficient models
- Local models
one contract · one audit trail · every model
How the router decides.
Routing is policy first, fit second. It keeps learning as models and prices change.
01
Task profiling
Each task carries its complexity, sensitivity and latency needs before a model is chosen.
02
Policy narrows the field
Data classification and residency rules decide which models are eligible at all.
03
Best fit wins
Among eligible models, the router balances intelligence, accuracy, cost and speed.
04
Measured and re-tuned
Outcomes feed back into the routes. As models and prices change, routing improves without rework.
What routing buys you
Observable per task, team and department
No lock-in
Models swap in and out behind one contract. Your workflows never notice.
Newest models, day one
When a better model ships, it joins the routing table. No migration project.
Spend under control
Expensive models only where they earn it. Spend controls and observability come built in.
Routing is one part of the layer.
Edrak Foundry pairs the router with grounded retrieval, context & memory and governance: one neutral layer between the enterprise and every model.
Explore Edrak Foundry →Explore Edrak Foundry →