Fee Models Across Networks: Reference Table

Entry last verified May 2026.

The models

Network type	Charged for	Price set by	Destination
Ethereum and compatible	Computational units consumed	Protocol base fee plus user tip	Base fee burned, tip to validator
Bitcoin	Transaction size in bytes	Fee market bidding	Entirely to the miner
Rollups	Execution plus data posting	Sequencer, tracking data costs	Sequencer, net of base layer costs
High-throughput layer 1	Varies, frequently near-fixed	Protocol parameters	Varies by protocol

The universal properties

The value transferred is not an input. Moving a small amount and a large amount cost the same. This is true everywhere and is the most persistent misunderstanding.

Failed operations still cost. Work performed is work charged.

Fees are paid in the network’s native asset. Holding a token without any of the fee asset means holding a balance you cannot move.

The Ethereum split

Since the 2021 fee upgrade, the fee has two parts.

The base fee is set by the protocol from recent block fullness and is destroyed rather than paid to anyone. It cannot be undercut; a transaction offering less is not included.

The priority fee is a tip that influences inclusion order. During quiet periods it is a small part of the total.

The size-based model

Bitcoin charges by transaction size. A transaction spending many small inputs is larger and therefore more expensive.

This is why consolidating small outputs during quiet periods saves money on every subsequent transaction, and it is an optimisation that does not exist on account-based chains.

The rollup composition

Rollup fees have three components: execution, data posted to the base layer, and the operator’s margin.

Since dedicated data capacity was introduced on Ethereum, the data component fell sharply and decoupled from base layer congestion. Rollup fees are consequently both low and stable.

What this means practically

Use a rollup for anything that does not require the base layer.

Keep native asset on every network where you hold tokens.

Batch where the model rewards it. Size-based models reward consolidation; metered models reward combining operations.

Time discretionary base layer activity to the weekly low, which is consistently the weekend morning window.

Venue charges versus network fees

A platform’s withdrawal fee is separate from the network cost. Some pass through, some charge a fixed amount with a margin.

Comparing the charge against the actual network fee on an explorer reveals which. Venues publishing the two separately, such as exchanges quoting a direct fiat pair, make the comparison possible without the arithmetic.