The Consensus Layer leverages a hybrid [[Proof of Intelligence (PoI)]] and [[Proof of Space (PoSp)]] model integrated with Substrate's BABE+GRANDPA consensus framework to achieve ledger agreement. In theoretical test environments with minimal latency and co-located validators, it can process up to **1,000 transactions per second (txs/s)**. This figure is derived using the formula TPS = 1 / (BlockTime + FinalizationTime), where BlockTime is set to 6 seconds (typical BABE block production) and FinalizationTime is approximately 1-2 seconds for GRANDPA finality, reflecting optimal conditions [109, 110]. However, in real-world production environments—considering factors such as network latency (e.g., 50ms round-trip time), geographic validator distribution, and [[ZKP/ZKP Base Layer/ZKP Blockchain/Technical Build Application Layer/Privacy-Preserving Computations with ZK Wrappers/Architecture and Workflow of ZK Wrappers|ZKP verification]] overhead—throughput is expected to range between **100-500 txs/s**. This adjusted estimate is based on empirical Substrate performance data, accounting for additional computational costs introduced by ZKP verification and custom pallet execution [118]. _**Note:** The 1,000 txs/s figure assumes negligible network latency and perfect validator synchronization; actual performance may be lower due to network delays, validator downtime, or increased transaction complexity, as reflected in the more conservative 100-500 txs/s range. It is important to distinguish between regular transaction throughput and ZK-verified AI computations. While our base layer can achieve 100-500 TPS for simple transactions, ZK-verified AI computations face computational constraints that limit their throughput relative to standard transactions. The inherent complexity of zero-knowledge verification for sophisticated AI workloads introduces verification overhead that must be carefully managed within the blockchain's throughput capacity._ _We are [[ZKP/ZKP Base Layer/ZKP Blockchain/Performance/Future Optimizations|actively exploring techniques]] to improve this throughput, including batch verification methods that amortize verification costs across multiple similar computations, and [[ZKP/Research/Privacy Pools for Scalability/Recursive Proofs for Batching|recursive proof systems]] that enable proof composition without linear verification scaling [47, 49]. These approaches represent promising directions for addressing the inherent tension between high-throughput consensus and computationally intensive ZK verification._ See also: [[ZKP/ZKP Base Layer/ZKP Blockchain/Performance/Application Layer|Application Layer]]