Model Routing & Cost Discipline
Route tasks to the right model tier to slash costs without sacrificing quality. The difference between a $500 and $5,000 monthly bill is smart routing.
Cost Routing Simulator
Drag each task into the model tier you think fits best. Watch the cost projection update in real-time. Can you beat the optimal routing?
Tier 1 — Fast & Cheap
Haiku 4.5 / GPT-5.6 Luna
$0.20 – $1 / M input tokens
Tier 2 — Balanced
Sonnet 5 / GPT-5.6 Terra
~$2 / M input tokens
Tier 3 — Premium
Opus 5 / GPT-5.6 Sol
$4 – $5 / M input tokens
There is now a fourth rung above these three: the frontier tier — Claude Fable 5.1 and GPT-6 Astra, both at $10 / M input tokens. Reserve it for the handful of tasks that genuinely need it; nothing in the simulator should be routed there by default.
Monthly Cost Projection
Pricing Calculator
Input your expected usage and see how multi-model routing compares to running everything on a single model.
Routing Split Assumption
Cost Optimisation Tips
Key Insight: The 80/15/5 Rule
In a well-designed agent system, roughly 80% of tasks are simple enough for the cheapest model tier — classification, routing, simple extraction, and formatting. Another 15% need a balanced model for synthesis and analysis. Only about 5% require premium reasoning power. Routing is the single biggest lever you have for controlling agent costs. A naive "send everything to the best model" approach can cost 10x more than intelligent routing with no meaningful quality difference on the simple tasks.