BNB Chain activated the Pasteur hard fork on BSC mainnet at 02:30 UTC on August 25, 2026. The upgrade shipped three protocol-level changes — BEP-675, BEP-682, and BEP-695 — targeting block execution capacity, cross-chain bridge verification, and validator key management. QANet benchmarks showed t...
BNB Chain activated the Pasteur hard fork on BSC mainnet at 02:30 UTC on August 25, 2026. The upgrade shipped three protocol-level changes — BEP-675, BEP-682, and BEP-695 — targeting block execution capacity, cross-chain bridge verification, and validator key management. QANet benchmarks showed throughput rising 88% from 1,237 to 2,324 transactions per second at the existing 450-millisecond block interval and 100 million gas limit.
The fork marks the opening move in BNB Chain's second-half 2026 roadmap, which targets a further doubling of mainnet throughput on a stated path toward a 10x improvement overall. A next-generation Layer-1 architecture targeting 100,000 TPS and sub-50ms pre-confirmations is planned for testnet by year-end, with mainnet deployment in early 2027.
Pasteur activated on BSC mainnet at block height corresponding to 02:30 UTC on August 25, 2026. Node operators were required to upgrade to client version v1.7.7 before the activation deadline. The Chapel testnet had been running a pre-release version (v1.7.6) since July 21, 2026, providing approximately five weeks of live testing data before mainnet deployment.
The upgrade delivered three BNB Evolution Proposals:
| BEP | Focus Area | Core Change | |-----|-----------|-------------| | BEP-675 | Block capacity | Pre-executed block submission; verification moved outside critical path | | BEP-682 | Bridge security | Duplicate validator entry rejection in cross-chain light-block verification | | BEP-695 | Validator governance | Key-rotation gap closure; slashing evasion prevention; delegated vote signer checks |
For end users and dApp developers, the fork required no action. Wallets, addresses, and balances were unaffected. The chain continued producing blocks at the 450-millisecond interval with no reported disruption.
BEP-675 is the throughput-focused proposal. It allows block builders to submit blocks that have already been executed. Under the previous architecture, validators performed execution and verification sequentially within the block production window. BEP-675 decouples these steps: validators check the block against consensus rules, sign, and broadcast it, then complete full verification afterward.
The result is more efficient utilization of the existing block space. QANet testing showed the following:
The mechanism works because the previous sequential processing left significant unused gas capacity in each block. By pre-executing transactions, builders can pack blocks closer to the gas limit without slowing block production.
BEP-682 addresses a specific vulnerability in BNB Chain's cross-chain bridge infrastructure. The pre-Pasteur verification process for cross-chain light blocks did not prevent the same validator from appearing multiple times in the approval list. A crafted request could count one validator's approval more than once, allowing a cross-chain transfer to pass with fewer independent validator approvals than the threshold required.
The fix is straightforward: the verification contract now rejects duplicate validator entries, ensuring the approval threshold reflects a genuine validator supermajority. Given that cross-chain bridge exploits have historically accounted for some of the largest losses in crypto — Ronin Network ($624 million, 2022), Wormhole ($320 million, 2022), and Nomad ($190 million, 2022) — the patch addresses a non-trivial attack vector.
The timing is notable. The Cosmos EVM vulnerability that surfaced in the same week (August 25, 2026) resulted in the TAC Chain losing approximately $7.5 million and KiiChain being drained of 148.3 million KII tokens across 18 exploit attempts, underscoring that bridge and cross-chain verification flaws remain an active threat category.
BEP-695 closes three governance and staking loopholes:
Key rotation rights revocation. When a validator replaces its operator key, the previous key now loses all management rights. Previously, old keys retained residual access, creating an operational security gap.
Slashing evasion prevention. Validators could previously escape pending slashing penalties by rotating their consensus keys before enforcement. Post-Pasteur, slashing and removal processes attach to the validator entity itself, not the operator address. Key rotation no longer resets the penalty clock.
Delegated vote signer verification. BNB Chain already blocked blacklisted addresses from voting directly in governance. However, restricted addresses could sign votes off-chain and have another address submit them. Pasteur adds a signer check to the governance contracts, verifying the original signer before counting a delegated vote.
The 88% throughput improvement is a QANet benchmark, not a confirmed mainnet result. BNB Chain's own documentation notes that "the reported performance gains came from controlled testing and are not guaranteed after mainnet activation." Network conditions on mainnet — variable transaction complexity, mempool congestion, geographic validator distribution — will affect real-world performance.
In H1 2026, BNB Chain's broader performance improvements (across multiple upgrades, not just Pasteur) achieved benchmark throughput of approximately 5,200 TPS, up from roughly 2,600 TPS at the start of the year. These figures represent controlled benchmarks rather than sustained mainnet load.
Mainnet daily transaction volume provides a more grounded measure. BNB Chain averaged approximately 12.4 million daily transactions post-H1 upgrades, with a single-day record of 17.6 million. The all-time high of 31 million daily transactions was reached on October 5, 2025.
| Metric | Value | Period | |--------|-------|--------| | All-time high daily transactions | 31 million | October 5, 2025 | | Average daily transactions (post-H1 upgrades) | 12.4 million | H1 2026 | | TVL growth | 40.5% YoY | 2025 | | Unique addresses | 700+ million | Cumulative | | Daily active users (BSC + opBNB) | 4+ million average | 2025 | | Stablecoin transactions processed | 5.3 billion+ | Since 2025 | | DEX market share (peak) | ~30% | 2025 | | Stablecoin annual growth | 133% YoY | 2025 | | RWA tokenized on-chain | $6.4 billion ATH | 2026 | | Gas price | 0.05 Gwei (down from 1 Gwei) | 2025 reduction | | Block time | 450ms (down from 3s) | Post-H1 2026 upgrades | | Finality | 1.125s (down from 7.5s) | Post-H1 2026 upgrades |
The gas price reduction — approximately 20x, from 1 Gwei to 0.05 Gwei — has been a significant driver of transaction volume growth, particularly for retail-oriented DeFi activity in Asian markets. PancakeSwap remains the ecosystem's anchor protocol.
Pasteur is the first of several planned upgrades in BNB Chain's H2 2026 pipeline. According to the official roadmap published in July 2026, the three stated priorities are speed, throughput, and protocol stability.
Near-term target (H2 2026): Double mainnet throughput again, continuing the path toward a cumulative 10x improvement. The H1 upgrades brought benchmark TPS from roughly 2,600 to 5,200. The H2 target implies reaching approximately 10,000+ TPS in benchmarks.
Next-generation L1 architecture (testnet late 2026, mainnet early 2027): A separate chain targeting 100,000+ TPS with sub-50ms pre-confirmations and sub-1-second block finality. Key architectural features include:
The next-generation L1 is explicitly designed for autonomous AI agents operating on-chain, positioning BNB Chain in the growing agentic infrastructure segment. The stated end-of-life throughput goal for the architecture is 1 million TPS.
BNB Chain's throughput trajectory exists within a broader Layer-1 performance race:
| Chain | Claimed/Benchmark TPS | Real-World Sustained TPS | Block Time | |-------|----------------------|--------------------------|------------| | BNB Chain (post-Pasteur) | 2,324 (QANet) / 5,200 (H1 benchmark) | ~1,500-2,000 est. | 450ms | | Solana | 65,000 (theoretical) | 1,000-4,000 (peak ~6,284) | 400ms | | Ethereum L1 | ~30 (theoretical) | ~15 | 12s | | Ethereum L1 + L2s | 5,000+ aggregate | Varies by rollup | Varies |
Solana maintains the highest real-world throughput among major L1s, processing roughly 1,000-4,000 TPS on a unified execution layer. BNB Chain's post-upgrade benchmarks place it in a comparable range under controlled conditions, though mainnet sustained load has not been independently verified at the 2,324 TPS level.
Ethereum's answer to the throughput question remains Layer-2 rollups, which aggregate to over 5,000 TPS but fragment liquidity across separate execution environments. BNB Chain, like Solana, pursues a monolithic scaling approach — pushing L1 throughput directly rather than offloading to rollups, though it also operates opBNB as an L2.
The Pasteur hard fork is operationally clean: no downtime, no user-facing changes, and three concrete security fixes alongside a throughput improvement. The bridge verification patch (BEP-682) is arguably the most consequential change, given the persistent threat of cross-chain exploits and the concurrent Cosmos EVM incident.
The throughput story is more nuanced. An 88% improvement in controlled testing is measurable progress, but the gap between benchmark capacity and real-world utilization on BNB Chain remains wide. The network's all-time transaction record of 31 million daily translates to an average of roughly 360 TPS — well below the 2,324 TPS QANet figure. Whether Pasteur enables the network to handle higher sustained loads during peak demand periods is the question that matters, and it can only be answered by mainnet data over time.
BNB Chain's broader ambition — a 100,000 TPS next-generation L1 by early 2027 — places it squarely in the race for institutional-grade throughput alongside Solana and Ethereum's rollup ecosystem. The economic question is whether throughput alone drives value capture, or whether it is a necessary-but-insufficient condition in a market where liquidity, developer ecosystem, and regulatory positioning carry equal weight.