[English Long Tweet] Brevis Research Report: zkVM and the Infinitely Verifiable Computation Layer of ZK Data Coprocessors

Chainfeeds Guide:

Brevis has built a multi-chain verifiable computation layer by integrating a general-purpose zkVM (Pico/Prism) with a data coprocessor (zkCoprocessor).

Source:

Jacob Zhao

Opinion:

Jacob Zhao: "Verifiable Computing" has become the general computation model for blockchains, with its core concept being "off-chain computation + on-chain verification." This paradigm allows blockchains to achieve nearly unlimited computational freedom while maintaining decentralization and trust-minimized security. Zero-knowledge proofs (ZKP) are the cornerstone of this system, with main application directions including scalability, privacy protection, and interoperability/data integrity. Among these, scalability was the earliest to be implemented, achieving high throughput and low-cost trust expansion by moving computation off-chain and verifying succinct proofs on-chain. The evolution path of ZK technology can be summarized as: L2 zkRollup → zkVM → zkCoprocessor → L1 zkEVM. L2 zkRollup first realized off-chain execution and on-chain verification, greatly improving performance; zkVM further expanded into a general-purpose verifiable computation layer, supporting cross-chain verification, AI inference, and cryptographic tasks; zkCoprocessor modularized this model, becoming a plug-and-play proof service for DeFi, RWA, and risk management; L1 zkEVM ultimately embeds zero-knowledge verification into the Ethereum execution layer, achieving real-time proving (RTP). This evolution represents the transformation of blockchain from "scalability" to "verifiability," ushering in the era of trustless computing. Ethereum's zkEVM journey has gone through two stages: the first stage (2022–2024), where L2 Rollup migrated the execution layer but introduced liquidity and state fragmentation; the second stage (2025–), where L1 RTP replaces N-of-N revalidation with 1-of-N proofs, improving throughput while maintaining decentralization. Beyond Ethereum, zero-knowledge computation is moving toward a broader field of verifiable computation, with core systems including zkVM and zkCoprocessor. zkVM (Zero-Knowledge Virtual Machine) is a general-purpose verifiable execution engine that supports running arbitrary programs on RISC-V, MIPS, or WASM instruction sets, with proof results verifiable on-chain, used for block validation, AI inference, and cross-chain tasks. Its advantages lie in generality and flexibility, but proof generation is costly and parallel optimization is complex. Representative projects include RiscZero, SuccinctLabs' SP1, and Brevis' Pico zkVM. In contrast, zkCoprocessor is more like a "verifiable module for specific scenarios," providing standardized computation and proof interfaces for DeFi, RWA, etc. Applications only need to call the SDK or API to obtain results and proofs, enabling fast integration and low cost, but with limited generality. Both share the "off-chain computation + on-chain verification" logic: the on-chain cost of executing computations is much higher than the combined cost of off-chain generation and on-chain verification. In business terms, zkVM follows a "Proving-as-a-Service" model, providing a computation engine for infrastructure; zkCoprocessor is "Proof-API-as-a-Service," serving the application layer with a SaaS structure that charges per task. The former builds a technical moat, while the latter drives commercial adoption. Together, they form the two poles of a trustless computation network: zkVM builds the computational foundation, and zkCoprocessor drives ecosystem prosperity. Brevis Network integrates the architectures of zkVM and zkCoprocessor to create a general-purpose, high-performance verifiable computation infrastructure, known as the "infinite computation layer." Its Pico zkVM adopts a modular architecture, decoupling the general execution layer from the hardware-accelerated coprocessor layer, achieving a "General + Specialized" dual-layer system, supporting multiple proof backends and recursive compression modules. Developers can use Rust to write business logic and automatically generate proofs, greatly lowering the entry barrier. Pico Prism further breaks through multi-GPU cluster performance, achieving an average proof time of 6.9 seconds and a 96.8% RTP coverage rate in a 64×RTX5090 environment, with a performance-cost ratio 3.4 times better than similar solutions. Brevis' zk Data Coprocessor gives smart contracts "memory," allowing them to access historical data and perform verification without trust, used for liquidity rewards, cross-chain identity, and data-driven DeFi. Its Incentra incentive layer achieves on-chain transparent distribution through ZK proofs and has already served protocols such as PancakeSwap, Euler, UsualMoney, and Linea. According to Brevis Explorer, by October 2025, the network has generated 125 millions ZK proofs, supported over $2.8 billions in TVL, and verified over $1 billions in transaction volume. Brevis, relying on the security layers of Ethereum and EigenLayer, has expanded to BNB, Linea, 0G, and other ecosystems, providing fundamental support for on-chain incentives, liquidity optimization, and cross-chain verification, building a flywheel-style verifiable computation network with two-way feedback between technology and applications. [Original text in English]