A LayerTree simulator puts the transaction for one member to leave a modeled Bitcoin (BTC) group of 1,048,576 at 581 virtual bytes (vB), using an opcode absent from Bitcoin proposals.
LayerTree's documentation describes the number as a simulator-measured model result and says the design does not currently run on Bitcoin's network. It is therefore not a reading from a live deployment.
The Oct. 15 draft by Lightning Labs developer Olaoluwa Osuntokun describes a shared Bitcoin output that commits to a tree of member balances held off-chain. To leave, a member would prove a balance and submit a transaction that updates the shared output. The tree has 65,536 positions at a depth of 16.
The 581-vB estimate uses the CCV tier with OP_MERKLESUMUPDATE; the same model estimates 541 vB for a 65,536-member tree. A different implementation, the FULL tier, puts a solo exit at 1,020 vB, or 1,030 vB when serialized. These figures rely on different script assumptions, so they do not measure the same implementation. The vB figure describes transaction size; the fee at broadcast depends on the prevailing rate per vB.
LayerTree's documentation says OP_MERKLESUMUPDATE has been specified and measured in a simulator, but has not been proposed for Bitcoin. Operators decide who can join. The design calls for members to check their tree branch before funding it because an invalid graft could leave a deposit inaccessible.
Every update spends the existing shared output and creates a replacement tied to the next tree state, so changes proceed one transaction at a time. Concurrent exits would compete to spend that output. Each confirmed exit changes the state available to the next member, and the final exit can remove the shared output.
A separate joining calculation for a group of about 1,000 estimates less than 1 vB per member if the operator covers the graft transaction and members supply their balances off-chain. That figure also measures transaction size, not a fixed payment.
LayerTree's verification page lists 7,690 script checks spanning four tiers, channel leaves and a Merkle-sum measurement, with implementation checks tallied separately. The project also provides a reference implementation, a simulator and tree rules checked in Lean. Its documentation gives no activation or rollout schedule for the required script changes.
