Selected model · —
from — est. — up to — / month ±50%
—
—
LLM tokens only; indexing, embeddings, storage, discounts are extra (Methodology below).
What the estimate assumes typical deployment
These are averages from live deployments, fixed so the estimate stays honest. Your traffic and adoption are the levers that matter.
Monthly cost by model at the traffic above
Pick the model your assistant runs on — one row per model family, fast tier to flagship. Platform-table rows are what Agent Studio bills; the rest are published list prices.
| Model | In $/Mtok | Out $/Mtok | Context | $ / month at this traffic |
|---|
Methodology — read this before quoting any number
Order-of-magnitude planning aids, not quotes. Three rules:
- Every constant is on this page; readers audit numbers instead of trusting them.
- Every figure is a ±50% band; deployments land inside it when the assumptions hold.
- A model without a defensible price shows no cost. Rows without a price count tokens; they refuse to invent dollars.
Price scopes. Platform rows use the exact per-1M table your bill comes from. List rows are published, dated prices. Auto-synced rows come from public registries.
Arithmetic mirrors the platform's estimation service; every price names its source; corrections go to data.js.