Monad in October 2026: the machine, the demand, and the number you should not quote

Public mainnet opened on 24 November 2025. By October 2026 the argument about Monad is no longer whether the client exists. It is which number on which page you are allowed to repeat. The introduction still says an Ethereum-compatible layer 1 with throughput of 10,000 transactions a second, a 300 millisecond block, and 600 millisecond finality. The newer why-Monad page is stricter. It puts a Basis column next to each figure. Throughput of 10,000 or more is Capacity. Block time of 300 milliseconds is Observed. Speculative finality of one slot and full finality of two slots are Protocol. A capacity is what the design can hold for typical transactions. An observation is what the live chain measured. A protocol figure is what the consensus rules say. Mixing the three is how a quiet Tuesday gets described as a broken chain, and how a marketing line gets described as a measurement.

The table, copied rather than improved

From the why-Monad page, as it stood when this essay was written: throughput 10,000 or more transactions a second, basis Capacity. Block time 300 milliseconds, basis Observed, with a note that the mean in September 2026 was about 302 milliseconds. Speculative finality one slot, 300 milliseconds, basis Protocol, and it can revert only if the original leader equivocates. Full finality two slots, 600 milliseconds, basis Protocol. Block gas budget 150 million gas, about 500 million gas a second, basis Observed on mainnet. Per-transaction gas limit 30 million, basis Protocol. If you need one of these for a contract, read the latest block and the current page. If you need one for an argument, cite the page, the basis, and the month. A blog that drops the basis has already chosen a side.

What 10,000 a second would actually look like

A slot is 0.3 seconds. Ten thousand transactions a second is 3,000 transactions in one block, if the transactions are the typical ones the capacity claim has in mind. A plain transfer is 21,000 gas. Three thousand of those are 63 million gas, which fits inside a 150 million gas block with room to spare. A swap that costs a few hundred thousand gas does not. Five hundred such swaps are already on the order of the whole block. The capacity claim is therefore a statement about a mix, not a promise that the chain will clear 10,000 DEX hops a second while also finalising them. The why-Monad page says this in the small print: the rate you see depends on how much gas each transaction uses. The small print is the claim.

DefiLlama, read on 27 October 2026, showed about 3.93 million transactions in the prior day. Divide by 86,400 seconds and you get about 45 transactions a second, averaged across the day. That is roughly two hundred times below the capacity slogan, and it is a perfectly ordinary reading of a chain that is not full. It is also an average. A burst inside one slot can be far above 45 and still far below 3,000. Quoting 45 as the maximum is as wrong as quoting 10,000 as the day. The adult sentence is: capacity on the order of ten thousand typical transactions a second, demand on a late-October day on the order of a few million transactions, concentrated however the venues concentrated it.

Ten thousand a second is a gas budget divided by a typical transaction. Forty-five a second is a day divided by a clock. Neither number refutes the other.

What the protocol page will stand behind

MonadBFT is proof of stake. The minimum block time in the consensus configuration is 300 milliseconds. Full finality is two slots. Speculative finality is one slot. Delegation is allowed. The same page says no other pipelined leader-based BFT combines speculative finality in one round, full finality in two, linear happy-path message complexity, and resistance to a tail fork, which is a leader throwing away the previous block. Happy path means no failures. Linear complexity is what lets the validator set be large rather than a handful of machines whispering. The client that runs this is Category Labs’ implementation, open source under GPL-3.0, consensus in one repository and execution in another, written in C++ and Rust. One open-source client is not the same as several independent clients. Say which fact you mean when you say decentralised.

Three market pages, three definitions

DefiLlama on 27 October 2026 put total value locked in Monad DeFi at about $979 million. Stablecoins were about $683 million, with USDC around a third of that stock. The same page showed about $22 million of DEX volume and about $20 million of perps volume in the prior day, about 8,900 active addresses, and about 1,700 new ones. The Foundation’s September highlights, published 1 October and dated to 30 September, said DeFi TVL above $1.01 billion, more than 4.3 million wallets, more than 860 million transactions since launch, more than $33 billion of cumulative DEX volume, and more than 170 apps. CertiK, writing on 20 October about figures as of 16 October, put TVL near $1 billion and described 100 percent uptime since public launch, more than 60 million uninterrupted blocks, a 99.903 percent successful proposal rate, and a 99.999 percent finalisation rate through 30 August 2026.

These are not three attempts at one number. A wallet that has ever touched the chain is not an address that transacted yesterday. Cumulative volume since November 2025 is not yesterday’s $22 million. TVL including a particular definition of deposits is not “total value including outstanding borrows,” which the Foundation put above $1.91 billion in the same September note. Adding deposits to borrows, or adding the stablecoin cap on top of TVL, produces a number nobody published. The honest paragraph names the page, the day, and the row. Near a billion dollars of DeFi TVL in October 2026 is the scale all three sources support. Anything sharper than that is a quotation, not a summary.

Who is running it

The September note says the network runs on consumer-grade machines, with on the order of 200 independently operated validators across more than 30 countries and more than 55 cities. That is the decentralisation claim the why-Monad page is organised around: performance without collapsing operators into a few data centres. You can pressure it. Count operators. Ask what share of stake the largest delegations control. Ask what a bug in the one production client does, given that the code is open and the implementation is still a family of clients rather than a crowd of them. You cannot pressure it by staring at a TPS slogan. Hardware that a person can buy is a real constraint on who may validate. It is not, by itself, a map of who holds the stake.

What to store, and what to throw away

Store the distinctions: capacity, observation, protocol, each with a date and a URL. Store that the mean block in September was about 302 milliseconds, so the live chain wobbles around the parameter rather than replacing it. Store that a day of a few million transactions does not fill the gas budget, so congestion is not the explanation when your swap reverts. The floor, the allowance, the pool, or the nonce is. Throw away the levels when a newer page exists. DefiLlama will have moved. The Foundation will publish another month. The block gas limit is a parameter and parameters move. The current-facts page is the one the docs tell agents to prefer over old blog numbers. When this essay and that page disagree, the page wins.