Routing playground
Pick a sample workload, tune the routing policy and see which model wins, and what it costs.
Prices, quality scores and latency are illustrative assumptions, not live benchmarks. Nothing is sent to an AI provider and nothing is charged.
Workload
Task
Request language
Optimize for
Sample request
Clasifica la urgencia de este ticket: Mi tarjeta empresarial fue rechazada y necesito pagar a un proveedor hoy.
Candidate models
Models below the quality bar are excluded before ranking by cost or latency.
| Model | Cost / request | Latency | Quality | Decision |
|---|---|---|---|---|
| Fast | $0.0075 | 0.7 s | 93 | Below bar |
| Balanced | $0.0080 | 1.2 s | 96 | Selected |
| Frontier | $0.0100 | 2.4 s | 98 | Eligible |
Estimated monthly spend: $570. Saved vs. direct: 43%. Clara discount: $30.
Where the baseline goes
43%
of the direct-provider baseline saved
- You pay$570
- Model routing$200
- Context & cache$100
- Language ramps$100
- Clara discount$30
- Total$1,000
Decision trace
Monthly cost breakdown
- Direct-provider baseline
- $1,000
- Model routing
- -$200
- Context & cache
- -$100
- Language ramps
- -$100
- Clara discount
- -$30
- Estimated monthly spend
- $570
Optimization of up to 40% depends on your workload and is not guaranteed. The Clara discount applies after optimization, so 40% and 5% combine to 43%, not 45%.
Ready to route real traffic?
Move one workload to Clara Router and compare it against your current provider.