> ## Documentation Index
> Fetch the complete documentation index at: https://docs.tied.com/llms.txt
> Use this file to discover all available pages before exploring further.
# Model costs and BYOK
> Model tokens pass through at exactly what the provider charged. Zero markup, at every tier — or bring your own key and Tied never touches the money.
Every Tied resolution uses one or more LLMs to route, reason and respond. **LLM cost is never bundled into the resolution price.**
Every invoice carries exactly two numbers: a flat orchestration fee per resolution, set by the resolution's class, and the LLM cost **at exactly what the provider charged — zero markup, at every tier**.
## What zero markup means mechanically
**No markup coefficient exists at any tier, at any value — including 1.0.** There is no field to set, no default to carry, and no floor to enforce. Billing mode is one of exactly two values: `itemized_passthrough` or `byok`.
The consequence for routing: Tied's revenue is identical whether a resolution runs on a frontier model or a cheap one, because the orchestration fee is set by the resolution's class and the token cost is passed through unchanged. Model selection has no effect on what Tied earns.
Tied publishes its model recommendations alongside the benchmarks behind them, so which model ran and why is always visible.
## Bring your own keys
You may supply your own provider keys — Anthropic, OpenAI, Google, Mistral and others — at **any tier, not just Enterprise**.
On your own provider account, your own enterprise agreement, your own negotiated rates.
It does not pass through our billing system, our balance sheet, or our invoice. It was never ours to
bill and never ours to waive.
The orchestration fee alone.
Bringing your own key **is** choosing your own model, which is why BYOK is available on every plan
while pinning a model from the Tied-hosted menu is an Advanced capability. See
[Plans](/pricing/plans).
## How tokens are priced on Tied-supplied keys
If you use Tied-supplied keys, we need a per-token price at the moment of each resolution — before the provider's monthly invoice exists.
Tied maintains its own price table for every model we route to. Each entry records the provider's
published rate and the date range it applies to. Resolutions are priced against the entry in effect at
the time they ran — so a mid-month provider price change does not retroactively rewrite your earlier
resolutions.
At month end we reconcile that table against the provider's actual invoice.
Any difference — in either direction — is allocated to customers in proportion to tokens consumed, and
appears as an explicit reconciliation line on your invoice.
The passthrough is not a promise about our intentions. It is an arithmetic identity that closes out
monthly, in public, on your bill.
## What appears on the receipt
Every resolution's receipt itemizes the LLM cost by **model, token count, provider rate and total** — separately from the orchestration fee, never blended into it.
A resolution that did not bill can still show an LLM cost. The two numbers are independent: the orchestration fee is what Tied charges for resolving something, and it is zero when nothing was resolved. The token cost is what the provider charged for tokens that were genuinely consumed. Zeroing out a real provider charge would make the receipt easier to read and less true.
## Model choice by plan
| | Starter | Business | Advanced |
| - | - | - | - |
| Multi-LLM least-cost routing (Tied picks) | — | ✓ | ✓ |
| Pin a specific Tied-hosted model, per intent or flow | — | — | ✓ |
| **Bring your own provider key** | **✓** | **✓** | **✓** |
## Next
The two-number model and the credit peg.
The seven classes and what each costs.
This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.