BNB Chain has published the first technical specifications of New L1, the provisional name of the EVM-compatible Layer 1 it plans to launch next to BNB Smart Chain. A documentation page dated 22 October 2026 sets a target of 200-millisecond blocks, preconfirmations in about 20 milliseconds and finality in under a second, with no public mempool and no new token: BNB stays the chain’s only gas, staking and governance asset. The chain runs today only on an internal devnet. The timetable, set in July and unchanged since, is a public testnet by the end of 2026 and mainnet in early 2027.
The project itself is not new. A fourth chain, next to BNB Smart Chain (BSC) and the two networks launched in 2023, opBNB and Greenfield, was first announced in its H2 2026 tech roadmap of 8 July. What changed this month is the level of detail: the July post gave headline targets, the October docs describe how blocks are built, how fees are charged and how BNB moves onto the new chain. Specialist outlets such as Crypto Briefing wrote the page up from 26 October, the day after Binance founder Changpeng Zhao commented on the plan on X.
Every figure below is a design target, dated to the document it comes from. None has been measured on a public network, because there is none yet.
What the BNB Chain New L1 docs add to the July roadmap
The 22 October page presents New L1 as “the next generation chain in the BNB Chain family”, aimed at high-frequency trading, real-time payments, confidential finance and activity driven by AI agents. BSC keeps its role as the broad, general-purpose chain. The new one is a clean-slate design for workloads that, according to the docs, outgrow BSC.
Two numbers do not line up with July, and readers comparing articles will see both. The 8 July roadmap promised preconfirmation in under 50 milliseconds; the 22 October docs say about 20 milliseconds. The July post also gave a throughput target of more than 100,000 transactions per second. The docs overview gives no TPS figure at all, so the 100,000 number that several outlets repeat this week comes from July, not from the new page. The July post gave no block time either: the 200 ms figure is new in October.
| Metric | Roadmap, 8 July 2026 | Docs, 22 October 2026 |
|---|---|---|
| Block interval | not given | 200 ms (target) |
| Preconfirmation | under 50 ms | about 20 ms |
| Finality | under 1 second | under 1 second |
| Throughput | 100,000+ TPS | not given |
| Public mempool | none | none |
| Testnet / mainnet | end 2026 / early 2027 | still ahead |
The timetable has not moved. “We plan to ship it on testnet by the end of 2026, with mainnet release following in early 2027,” the July roadmap said. The docs only say that the public testnet and mainnet are still ahead.
The design: 200 ms blocks, 20 ms preconfirmations and no public mempool
According to the 22 October docs, consensus is an “enhanced” version of Parlia, the proof-of-staked-authority system BSC already uses, with each validator producing 16 blocks in a row during its turn. The target is a block every 200 milliseconds. Inside each block, smaller “SubBlocks” let the producer confirm a transaction to the sender in about 20 milliseconds. Finality, the point after which the order of transactions cannot be reversed, is meant to come in under a second through BLS signature votes collected across blocks.
A preconfirmation, the documentation stresses, is a fast signal from the block producer and not finality. Execution is only final once a later block records an “ExecutionCommitment” and that block is finalized in turn. An app that shows a trade as done at 20 ms is relying on a promise that can still be withdrawn.
New L1 has no global mempool, the public waiting room where pending transactions sit on Ethereum or BSC and where bots spot trades to front-run. Transactions go straight to the current block producer and the next one. The July roadmap called this routing TxStream. David Z, the chain’s chief technology officer, told The Block on 8 July that “TxStream doesn’t eliminate MEV. Nothing does. It makes the dominant attacks impractical by design.”
Fees work differently too. Users pay for the gas limit they declare, not the gas their transaction ends up using: block space is sold on declared gas before execution, so padding the limit costs money. A “Multi-Lane” system reserves gas capacity per class of transaction, so a burst of trading is not supposed to crowd out oracle updates or liquidations. The July roadmap described the same idea under the name PriorityLane, with lanes for oracles, liquidations and bridges.
Under the hood, execution runs on proof of stake coin with a protocol that stakes it for you and issues a transferable token (such as Lido stETH or rETH) representing the deposit plus rewards, which you can use or sell while the coins stay staked.">reth and revm, the Rust Ethereum client and virtual machine. State is held in a flat key-value store with an LtHash commitment instead of Ethereum’s Merkle-Patricia trie. That choice has a cost for developers: the docs list eth_getProof, EIP-4844 blobs and EIP-7702 among features that are not supported, so “EVM-compatible” does not mean every Ethereum tool works unchanged.
Two features target users directly. Gas sponsorship through native account abstraction (a dedicated transaction type, 0x76) lets apps pay fees for their users, with signing by passkey and access keys limited to set actions. An opt-in shielded smart contract holding reserves of two or more tokens, supplied by users, that a decentralized exchange uses to execute trades automatically at a price set by a formula.">pool hides sender, recipient and amount for those who choose it. The overview does not say how that pool will fit with the anti-money-laundering checks exchanges apply.
BNB stays the only token, and the bridge is not finished
No new coin comes with it. BNB is the native gas asset of New L1, bridged over from BSC. Staking BNB through the StakeHub contract mints govBNB one for one, a non-transferable token that only measures governance weight. Governance is token-weighted, with a timelock before changes apply.
Zhao welcomed that choice in a post on X on 25 October, according to Cointribune; outlets summarise it as “multiple chains, one coin”, with slightly different wording from one to the next. We could not open the post to check its exact text, so we do not quote it. In practice the design sends gas fees and staking on the new chain to BNB rather than to a second token, the same coin whose supply is cut each quarter, most recently in the 37th quarterly burn of 1.65 million BNB. “The goal is growing the total BNB Chain pie, not re-slicing it,” David Z told The Block in July.
The bridge is not ready. The native link between BSC and New L1 is, per the docs, “still in development”. Until it ships, BNB cannot move to the new chain in the way the design assumes, and how BSC liquidity and fees will interact with a faster sibling is an open question that neither the roadmap nor the docs answer.
Markets have not moved on it. BNB traded near $738.6 at 11:32 UTC on 27 October, down 0.27% over 24 hours and up 1.63% over seven days, according to Coindaro’s market snapshot: no reaction to the docs was visible at that time.
Where BSC stands today, after the Fermi fork
The new chain will sit next to a BSC that has already been sped up this year. The Fermi hard fork went live on BSC mainnet at 02:30 UTC on 14 January 2026, the time set in BNB Chain’s blog post of 7 January, and cut the block interval from 750 to 450 milliseconds. Block timestamps on BSC show it: an average of 750 ms between blocks in the hours before, 450 ms after. The July roadmap then counted the first-half results, from January to June: block interval 750 to 450 ms, in-memory finality 1,125 to 650 ms, and benchmark throughput from about 2,800 to about 5,200 TPS. For the second half, it set “a 2x throughput increase on BSC mainnet” as the immediate objective.
So the gap the new chain is meant to open is large on paper: blocks more than twice as fast as BSC’s 450 ms, and, if the July target holds, roughly 20 times BSC’s benchmark throughput. The comparison sets a lab benchmark against a target, not two measurements.
For a reader new to the distinction, New L1 is a separate base chain with its own validators and consensus, not a layer 2 that posts its data to BSC. Our explainer on the difference between a layer 1 and a layer 2 covers why that matters for security and fees.
Developers get familiar tools. The docs say New L1 uses the same tooling as BSC, opBNB and Greenfield, but the gaps listed above (no eth_getProof, no EIP-4844 or EIP-7702 transactions) mean contracts and wallets built for Ethereum need testing before they move.
What is still unknown before the 2026 testnet
The docs themselves flag several gaps.
- The chain has no final name; “New L1” is a placeholder.
- Neither a throughput figure nor a validator count appears in the docs overview, so the size of the validator set is unconfirmed.
- There is no public testnet, block explorer or RPC endpoint, so outside developers cannot test any of the numbers.
- The native BSC bridge is unfinished, and mainnet is dated only “early 2027”.
- Preconfirmations can be withdrawn, and the overview does not say what apps should show users in that case.
David Z was also frank about limits in July. On high-frequency trading he said: “For co-located HFT, a CEX still wins today. For everyone else, this is the CEX experience without the custodial risk.” On post-quantum security, which BNB Chain is testing as a hybrid layer: “Nobody in the industry has a complete quantum migration scheme yet, including us.”
The next checkpoint is the public testnet, due by the end of 2026 under the July plan. It would be the first time the 200 ms and 20 ms targets meet traffic from outside developers.