Cheapest network to send right now
Data: your browser talks directly to public chain nodes, mempool.space, Blockchair and CoinGecko. Your browser reads gas and fee figures from public chain nodes and public fee indexes, and coin prices from CoinGecko, so each of those operators sees your IP address while the page is open. There is nothing to type here, so what you were about to send never reaches us.
Twelve networks, priced this minute, cheapest first. There is nothing to fill in - the page opens with the answer already on screen.
A token transfer is not a coin transfer, and the gap is about threefold
One list shows what it costs to send USDT on the eight chains that carry it; the other adds Bitcoin, Litecoin, Dogecoin and Bitcoin Cash and compares what it costs to move value at all. Sending a chain’s own coin uses a fixed 21,000 gas. Sending USDT is a contract call: it has to write a balance into the token’s own storage, which runs about 65,000 gas on an ordinary EVM chain. Pages that price a stablecoin send at 21,000 understate it roughly threefold, and that is the single most common error in this category. The USDT list here uses token gas, the any-coin list uses native gas, and each row prints what it assumed.
- Native send: 21,000 gas, a protocol constant rather than an estimate.
- USDT send to somebody who already holds it: about 65,000 gas.
- USDT send to an address that has never held it: about 82,100 gas, because the recipient’s balance goes from zero to non-zero and that storage write costs far more than an update.
The first send to a new address costs more everywhere
Every row carries a second figure for the first time value reaches an address that has never held the thing you are sending, because the surcharge is not a rounding error. On TRON a USDT transfer to a fresh recipient consumes 130,285 energy against 64,285 for one who already holds it - more than double, and a page that prints a single averaged number is wrong for whichever half of its readers it did not describe. On Solana the recipient needs a token account, whose rent-exempt deposit is about 0.00204 SOL; that deposit is recoverable if the account is ever closed, which is why it is shown apart from the fee rather than added into it.
Where each number comes from, and which ones are measured
Two kinds of figure appear here and the difference is stated on every row. A measured median is what people actually paid in the last twenty-four hours, which is a stronger answer than any arithmetic: Litecoin, Dogecoin and Bitcoin Cash are read that way. Everything else is computed - the live gas or fee rate multiplied by a stated amount of work, then by the live coin price. Bitcoin uses the next-block rate over a 141-vbyte spend, the ordinary one-input, two-output transaction.
- A rollup pays twice: once to run the transaction on itself, and again to publish it on Ethereum. The second half is invisible in a rollup’s own gas price, and on a quiet layer 2 it can be more than ten times the first half - so a page that shows execution only ranks that chain far cheaper than it is. Both halves are read here, each from the rollup’s own on-chain fee oracle.
- Costs round up. A figure here is never lower than the real one, because funding a wallet with exactly the amount a page promised and watching the send fail is the expensive direction to be wrong in.
- Amounts are handled as whole base units - wei, sun, lamports, satoshi - and become a decimal only when printed.
- A network whose figure could not be read keeps its place in the list, says so, and sorts to the bottom rather than the top.
Cheapest is not the same as correct
The list answers what a transfer costs, not where to send. A cheaper network is only useful if whoever is receiving the funds actually supports it, and a correct address on the wrong network is the most expensive way anyone has ever saved on fees. Confirm the asset, the network and the address together, in that order, before anything moves. The confirmation time on each row is one block at that chain’s own target spacing; anyone crediting your funds will wait for several.
What this page does not do
It reads once when it loads and stamps the time. There is no polling until you switch auto-read on, because a fee page that hammers public sources on a timer is how those sources stop answering for everybody. Switching between the two lists costs no requests at all - both are computed from the same readings. And nothing is sent to us: there is no address field, no amount field, and no account.
Questions
- Why is your USDT figure so much higher than a gas tracker’s?
- Because a gas tracker usually prices a plain coin transfer at 21,000 gas. A USDT transfer is a contract call that writes into the token’s storage and costs roughly 65,000 gas - about three times as much - and more still the first time it reaches an address that has never held the token.
- Why do you show two numbers for TRON?
- Because the network charges by work done, and writing a USDT balance for an address that never had one is far more work. Measured on the live network, that is 130,285 energy against 64,285 - more than double. Averaging the two would misprice whichever case you are actually in.
- One of the rows says the price could not be read. Is the page broken?
- No. Each network is read separately, so one unreachable source affects one row. That row keeps its place, says what happened and sorts to the bottom, and the other rows still answer the question. Read again in a moment and it usually returns.
- Are these the fees I would pay on an exchange?
- No. This is what the network itself charges to move the funds. Any service that sends on your behalf prices its own costs into the rate it quotes you, which is settled by a quote with an expiry rather than by a fee table.