Brevis Research Report: The Infinite Trustworthy Computing Layer of ZKVM and Data Co-Processors

Author: Jacob Zhao Source: mirror

The paradigm of “off-chain computation + on-chain verification” for Verifiable Computing has become the general computing model for blockchain systems. It allows blockchain applications to achieve almost infinite computational freedom while maintaining decentralization and trustlessness security. Zero-Knowledge Proofs (ZKP) are the core pillar of this paradigm, with applications primarily focused on three fundamental areas: Scalability, Privacy, and Interoperability & Data Integrity. Among these, scalability is the earliest scenario where ZK technology has been implemented, achieving high TPS and low-cost trusted scalability by moving transaction execution off-chain and verifying results on-chain with short proofs.

The evolution of ZK trusted computing can be summarized as L2 zkRollup → zkVM → zkCoprocessor → L1 zkEVM. Early L2 zkRollup migrated execution to layer two and submitted validity proofs on layer one, achieving high throughput and low-cost scaling with minimal changes. zkVM then expanded into a general verifiable computing layer, supporting cross-chain verification, AI inference, and cryptographic computation (representative projects: Risc Zero, Succinct, Brevis Pico). zkCoprocessor developed in parallel as a scenario-based verification module, providing plug-and-play computing and proof services for DeFi, RWA, risk control, etc. (representative projects: Brevis, Axiom). In 2025, the concept of zkEVM extended to L1 real-time proving (Realtime Proving, RTP), building verifiable circuits at the EVM instruction level, integrating zero-knowledge proofs directly into Ethereum's mainnet execution and verification process, becoming a natively verifiable execution mechanism. This context reflects the technological leap of blockchain from “scalable” to “verifiable,” opening a new phase of trusted computing.

1. The Road to Ethereum zkEVM Scaling: From L2 Rollup to L1 Real-Time Proof

The zkEVM scaling path of Ethereum goes through two stages:

• Phase One (2022–2024): L2 zkRollup will execute transfers to the second layer, submitting validity proofs on the first layer; significantly reducing costs and increasing throughput, but leading to liquidity and state fragmentation, while L1 is still constrained by N-of-N re-execution.
• Stage Two (2025–): L1 Real-time Proving (RTP) replaces re-execution with “1-of-N proof + lightweight validation across the network,” enhancing throughput without sacrificing decentralization, and is still under evolution and development.

L2 zkRollup Phase: Balancing Compatibility and Scalability Performance

In 2022, during the flourishing phase of Layer 2 ecology, Ethereum founder Vitalik Buterin proposed four categories of ZK-EVM (Type 1–4), systematically revealing the structural trade-offs between compatibility and performance. This framework established a clear coordinate for the subsequent zkRollup technology roadmap:

• Type 1 Fully Equivalent: Consistent with Ethereum bytecode, minimal migration cost, slowest proof. Taiko.
• Type 2 Fully Compatible: Very few underlying optimizations, strongest compatibility. Scroll, Linea.
• Type 2.5 quasi-compatible: minor changes (gas/pre-compilation, etc.) for performance. Polygon zkEVM, Kakarot.
• Type 3 partial compatibility: larger changes, can run most applications but difficult to fully reuse L1 infrastructure. zkSync Era.
• Type 4 Language Level: Abandon bytecode compatibility, directly compile from high-level languages to circuits, optimal performance but requires ecosystem reconstruction (Representative: Starknet / Cairo).

The current L2 zkRollup model has become mature: by migrating execution to layer two and submitting validity proofs on layer one, it minimally adapts to the Ethereum ecosystem and toolchain, becoming a mainstream solution for scalability and cost reduction. Its proof objects are L2 blocks and state transitions, while settlement and security remain anchored on L1. This architecture significantly improves throughput and efficiency, while maintaining high compatibility for developers, but it also brings liquidity and state fragmentation, and L1 is still limited by the N-of-N re-execution bottleneck.

L1 zkEVM: Real-time proof reshaping Ethereum light verification logic

In July 2025, the Ethereum Foundation published an article titled “Shipping an L1 zkEVM #1: Realtime Proving,” officially proposing the L1 zkEVM roadmap. The L1 zkEVM upgrades Ethereum from N-of-N re-execution to 1-of-N proof + fast verification across the network: a few provers generate short proofs for the entire EVM state transition, with all validators only performing constant-time verification. This solution achieves L1 level real-time proving without sacrificing decentralization, enhances the mainnet's Gas limit and throughput, and significantly lowers the hardware threshold for nodes. The implementation plan is to replace traditional execution clients with zk clients, running in parallel initially, and gradually becoming the new norm at the protocol layer once performance, security, and incentive mechanisms mature.

• N of N Old Paradigm: All validators execute the entire block of transactions to verify it, which is secure but has limited throughput and high peak fees.
• 1 of N New Paradigm: A small number of provers execute the entire block and produce short proofs; the entire network only performs constant time verification. The verification cost is much lower than re-execution, which can safely increase the L1 gas limit and reduce hardware requirements.

L1 zkEVM roadmap three main lines

1. Realtime Proving: Completing the full block proof within a 12-second slot time, compressing delays through parallelization and hardware acceleration;
2. Client and protocol integration: standardized proof verification interface, optional first, then default;
3. Incentives and Security: Establish the Prover market and fee model to strengthen anti-censorship and network activity.

Ethereum L1 Real-Time Proof (RTP) uses zkVM to re-execute entire blocks of transactions off-chain and generate cryptographic proofs, allowing validators to verify a small proof within 10 seconds without recalculating, thereby achieving “verification instead of execution”, significantly enhancing Ethereum's scalability and trustless verification efficiency. According to the official zkEVM Tracker page of the Ethereum Foundation, the main teams currently participating in the L1 zkEVM real-time proof route include SP1 Turbo (Succinct Labs), Pico (Brevis), Risc Zero, ZisK, Airbender (zkSync), OpenVM(Axiom, and Jolt(a16z).

2. Beyond Ethereum: Universal zkVM and zkCoprocessor

Outside of the Ethereum ecosystem, Zero-Knowledge Proof (ZKP) technology has also extended into the broader field of Verifiable Computing, forming two technical systems centered around zkVM and zkCoprocessor.

zkVM: Universal Verifiable Computing Layer

A verifiable execution engine for arbitrary programs, common instruction set architectures include RISC-V, MIPS, and WASM. Developers can compile business logic to zkVM, which is executed off-chain by a prover to generate zero-knowledge proofs (ZKP) that can be verified on-chain, applicable for Ethereum L1 block proofs, cross-chain verification, AI inference, cryptographic computations, and complex algorithms. Its advantages lie in its versatility and broad adaptability, but it has complex circuits and high proof costs, requiring multi-GPU parallelism and strong engineering optimization. Representative projects include Risc Zero, Succinct SP1, Brevis Pico / Prism.

zkCoprocessor: Scenario-based Verifiable Module

Provide “plug-and-play” computing and proof services for specific business scenarios. The platform pre-configures data access and circuit logic (such as reading historical on-chain data, TVL, yield settlement, identity verification, etc.), and application parties can obtain computation results and proofs for on-chain consumption through SDK/API calls. This model is quick to get started, performs well, and is low-cost, but has limited versatility. Typical projects include Brevis zkCoprocessor, Axiom, etc.

Overall, zkVM and zkCoprocessor both follow the trusted computing paradigm of “off-chain computation + on-chain verification,” verifying off-chain results on-chain through zero-knowledge proofs. Its economic logic is based on the premise that the cost of direct execution on-chain is far higher than the combined cost of off-chain proof generation and on-chain verification.

The key difference between the two in terms of versatility and engineering complexity is:

• zkVM is a general-purpose computing infrastructure suitable for complex, cross-domain, or AI scenarios, offering the highest flexibility.
• zkCoprocessor is a modular verification service that provides low-cost, directly callable verification interfaces for high-frequency reusable scenarios (DeFi, RWA, risk control, etc.).

In the commercial path, the difference between zkVM and zkCoprocessor is:

• zkVM adopts the Proving-as-a-Service model, charging per proof (ZKP), primarily targeting infrastructure clients such as L2 Rollup. It is characterized by large contract sizes, long cycles, and stable gross margins;
• zkCoprocessor focuses on Proof API-as-a-Service, charging based on tasks through API calls or SDK integration, which is closer to a SaaS model, targeting application layer protocols like DeFi, with fast integration and strong scalability.

Overall, zkVM is the underlying engine for verifiable computation, while zkCoprocessor is the application layer verification module: the former builds a technological moat, and the latter drives commercial implementation, together forming a general-purpose trusted computing network.

3. The Product Landscape and Technical Path of Brevis

Starting from Ethereum's L1 Realtime Proving, ZK technology is gradually moving towards a verifiable computing era centered around the general zkVM and zkCoprocessor architecture. Brevis Network is a fusion of zkVM and zkCoprocessor, building a general verifiable computing infrastructure centered on zero-knowledge computation, combining high performance and programmability — The Infinite Compute Layer for Everything.

( 3.1 Pico zkVM: A Modular Proof Architecture for General Verifiable Computation

In 2024, Vitalik proposed the "General Execution Layer + Coprocessor Acceleration Layer" (glue & coprocessor) architecture in “Glue and Coprocessor Architectures.” Complex computations can be divided into general business logic and structured intensive computing—the former pursues flexibility (like EVM, Python, RISC-V), while the latter pursues efficiency (like GPU, ASIC, hash modules). This architecture is becoming a common trend in blockchain, AI, and cryptographic computing: EVM speeds up through precompiles, AI leverages GPU parallelism, and ZK proofs combine general VMs with dedicated circuits. The key for the future is to optimize the “glue layer” for security and development experience, while the “coprocessor layer” focuses on efficient execution, achieving a balance between performance, security, and openness.

![])https://img-cdn.gateio.im/webp-social/moments-c46288543a83af189338fd6a8ba5d1ef.webp###

Pico zkVM, developed by Brevis, is a representative implementation of this concept. Through the “general zkVM + co-processor acceleration” architecture, it combines flexible programmability with high-performance computing of dedicated circuits. Its modular design supports multiple proof backends (KoalaBear, BabyBear, Mersenne31) and allows for the free combination of execution, recursion, compression, and other components to form ProverChain.

The modular system of Pico not only allows for the free reorganization of core components but also enables the introduction of new proof backends and application-level co-processors (such as on-chain data, zkML, and cross-chain verification), achieving continuously evolving scalability. Developers can directly use the Rust toolchain to write business logic, automatically generating cryptographic proofs without requiring a background in zero-knowledge, significantly lowering the development threshold.

Compared to the relatively monolithic RISC-V zkVM architecture of Succinct SP1 and the general RISC-V execution model of RISC Zero R0VM, Pico achieves decoupling and scalability of execution, recursion, and compression phases through a Modular zkVM + Coprocessor System, supporting multi-backend switching and coprocessor integration, thus forming differentiated advantages in performance and scalability.

( 3.2 Pico Prism: Performance Breakthrough of Multi-GPU Clusters

Pico Prism is a significant breakthrough by Brevis on a multi-server GPU architecture, setting new records under the Ethereum Foundation's “Real-Time Proving (RTP)” framework. Achieving an average proof time of 6.9 seconds with a 96.8% RTP coverage rate on a 64×5090 GPU cluster, it ranks first among its peers in zkVM performance. The system has been optimized at the architectural, engineering, hardware, and system levels, marking a transition of zkVM from a research prototype to production-grade infrastructure.

1. Architecture Design: Traditional zkVMs (such as SP1, R0VM) primarily rely on single-machine GPU optimization. Pico Prism is the first to achieve multi-server, multi-GPU cluster parallel proving (Cluster-Level zkProving), expanding zk proofs into a distributed computing system through multi-threading and sharding scheduling, significantly enhancing parallelism and scalability.
2. Engineering Implementation: Build a multi-stage asynchronous pipeline (Execution / Recursion / Compression) and cross-layer data reuse mechanism (proof chunk caching and embedding reuse), supporting multiple backend switching (KoalaBear, BabyBear, M31), significantly improving throughput efficiency.
3. Hardware Strategy: With a configuration of 64×RTX 5090 GPUs (approximately $128K), Pico Prism achieves an average proof time of 6.0–6.9 seconds, a 96.8% RTP coverage rate, and an improvement in performance/cost ratio of about 3.4 times, outperforming SP1 Hypercube (160×4090 GPUs, 10.3 seconds).
4. System Evolution: As the first zkVM that meets the Ethereum Foundation RTP criteria (>96% sub-10s, <) cost), Pico Prism marks the transition of zk proof systems from research prototypes to mainnet-level production infrastructure, providing more cost-effective zero-knowledge computing solutions for scenarios such as Rollup, DeFi, AI, and cross-chain verification.

3.3 ZK Data Coprocessor: Blockchain Data Intelligent Zero-Knowledge Coprocessing Layer

The “lack of memory” in the native design of smart contracts — the inability to access historical data, recognize long-term behavior, or perform cross-chain analysis. Brevis provides a high-performance Zero-Knowledge Coprocessor (ZK Coprocessor), offering smart contracts access to cross-chain historical data and trusted computing capabilities, validating and computing all historical states, transactions, and events on the blockchain, applied in scenarios such as data-driven DeFi, active liquidity management, user incentives, and cross-chain identity recognition.

The workflow of Brevis consists of three steps:

1. Data Access: Smart contracts read historical data trustlessly through APIs;
2. Calculation Execution: Developers define business logic using the SDK, which is then calculated off-chain by Brevis to generate ZK proofs.

Result verification: The proof result is returned on-chain, verified by the contract and calls subsequent logic.

![]$100K https://img-cdn.gateio.im/webp-social/moments-34c4b4cdde3635acaed075831947e4ae.webp###

Brevis supports both Pure-ZK and CoChain (OP) models: the former achieves complete trust minimization but at a higher cost; the latter allows for verifiable computation at a lower cost through PoS validation and ZK challenge mechanisms. Validators stake on Ethereum, and if the result is challenged successfully by ZK proof, they will be penalized, thus achieving a balance between security and efficiency. Through the integration of ZK + PoS + SDK architecture, Brevis strikes a balance between security and efficiency, building a scalable trusted data computation layer. Currently, Brevis has served protocols such as PancakeSwap, Euler, Usual, Linea, etc., all zkCoprocessor collaborations are based on Pure-ZK model, providing trusted data support for DeFi, reward distribution, and on-chain identity systems, enabling smart contracts to genuinely possess “memory and intelligence.”

( 3.4 Incentra: ZK-based “Verifiable Incentive Distribution Layer”

Incentra is a trusted incentive distribution platform powered by Brevis zkCoprocessor, providing secure, transparent, and verifiable reward calculation and distribution mechanisms for DeFi protocols. It directly verifies incentive results on-chain through zero-knowledge proofs, achieving trustless, low-cost, and cross-chain incentive execution. The system completes reward calculation and verification in ZK circuits, ensuring that any user can independently verify the results; it also supports cross-chain operations and access control, enabling compliant and secure automated incentive distribution.

Incentra mainly supports three types of incentive models:

• Token Holding: Long-term holding rewards calculated based on ERC-20 time-weighted average balance (TWA);
• Concentrated Liquidity: Distributes liquidity rewards based on the fee ratio of AMM DEX, compatible with ALM protocols such as Gamma and Beefy.
• Lend & Borrow: Calculate lending rewards based on the average balance and debt.

The system has been applied to projects such as PancakeSwap, Euler, Usual, and Linea, achieving a fully trusted closed loop from incentive calculation to distribution, providing ZK-level verifiable incentive infrastructure for DeFi protocols.

) 3.5 Brevis Product Technology Stack Overview

![]###https://img-cdn.gateio.im/webp-social/moments-52317d5796169cb68f9a65b545ce3b4c.webp###

4. Brevis zkVM Technical Indicators and Performance Breakthroughs

The L1 zkEVM real-time proving standard (Realtime Proving, RTP) proposed by the Ethereum Foundation (EF) has become the industry consensus and entry threshold for whether zkVM can enter the Ethereum mainnet verification route, with core evaluation indicators including:

• Latency Requirement: P99 ≤ 10 seconds (matching Ethereum's 12 seconds block time);
• Hardware constraints: CAPEX ≤ $100K, power consumption ≤ 10kW (suitable for home/small data center);
• Security Level: ≥128-bit (Transition Period ≥100-bit);
• Proof size: ≤300 KiB;

System requirements: Must not rely on trusted settings; core code must be fully open source.

In October 2025, Brevis released the report “Pico Prism — 99.6% Real-Time Proving for 45M Gas Ethereum Blocks on Consumer Hardware”, announcing that its Pico Prism became the first zkVM to fully pass the Ethereum Foundation (EF) Real-Time Block Proving (RTP) standard.

With a configuration of 64×RTX 5090 GPUs (approximately $128K), Pico Prism achieved an average latency of 6.9 seconds, with 96.8% of transactions under 10 seconds and 99.6% under 12 seconds in a 45M gas block, significantly outperforming Succinct SP1 Hypercube (36M gas, average time 10.3s, with 40.9% under 10 seconds). Under conditions of a 71% reduction in latency and halved hardware costs, the overall performance/cost efficiency improved by approximately 3.4×. This achievement has been publicly recognized by the Ethereum Foundation, Vitalik Buterin, and Justin Drake.

5. Brevis Ecosystem Expansion and Application Implementation

The ZK Coprocessor of Brevis ( zkCoprocessor ) is responsible for handling complex calculations that dApps cannot efficiently complete (such as historical behavior, cross-chain data, and aggregation analysis) and generating verifiable zero-knowledge proofs (ZKP). Only a small proof needs to be verified on-chain to securely call the result, significantly reducing gas, latency, and trust costs. Compared to traditional oracles, Brevis provides not just “results,” but also “mathematical guarantees of result correctness,” and its main application scenarios can be divided into the following categories.

• Intelligent DeFi: Achieving intelligent incentives and differentiated experiences based on historical behavior and market conditions (PancakeSwap, Uniswap, MetaMask, etc.)
• RWA and Stable Token Growth: Automated allocation of stablecoin and RWA yields through ZK verification (OpenEden, Usual Money, MetaMask USD)
• Privacy Decentralized Exchange (DEX with Dark Pools): A privacy trading model that uses off-chain matching and on-chain verification, will be launched soon.
• Cross-chain Interoperability: Supports cross-chain re-staking and Rollup–L1 interoperability, building a shared security layer (Kernel, Celer, 0G)
• Blockchain Bootstrap: Using ZK incentive mechanisms to support the cold start and growth of new public chain ecosystems (Linea, TAC)
• High-performance public chains (100× Faster L1s): Enhancing the performance of public chains like Ethereum and BNB Chain through Real-Time Proof (RTP) technology.

Verifiable AI: Integrating privacy protection and verifiable reasoning to provide trustworthy computing power for AgentFi and the data economy (Kaito, Trusta)

According to Brevis Explorer data, as of October 2025, the Brevis network has generated over 125 million ZK proofs, covering nearly 95,000 addresses and 96,000 application requests, widely serving scenarios such as reward distribution, transaction verification, and staking proof. On the ecological level, the platform has distributed incentives totaling approximately 223 million USD, with a supported TVL exceeding 2.8 billion USD, and the related transaction volume has surpassed 1 billion USD.

Currently, Brevis's ecological business mainly focuses on two directions: DeFi incentive distribution and liquidity optimization. The core power consumption is contributed by four projects: Usual Money, PancakeSwap, Linea Ignition, and Incentra, which together account for over 85%. Among them,

• Usual Money (46.6M proofs): Demonstrates its long-term stability in large-scale incentive distribution;
• PancakeSwap (20.6M): Demonstrates Brevis's high performance in real-time fee and discount calculations;
• Linea Ignition (20.4M): Validate its high concurrency processing capabilities in L2 ecosystem activities;
• Incentra (15.2%): Marks the evolution of Brevis from SDK tools to a standardized incentive platform.

In the field of DeFi incentives, Brevis relies on the Incentra platform to support multiple protocols for transparent and continuous reward distribution:

• Usual Money annual incentive scale exceeds $300M, providing continuous profits for stablecoin users and LPs.
• OpenEden and Bedrock implement the distribution of US Treasury and Restaking yields based on the CPI model;
• Protocols like Euler, Aave, and BeraBorrow use ZK verification for loan position and reward calculations.

In terms of liquidity optimization, PancakeSwap, QuickSwap, THENA, Beefy, and others utilize Brevis's dynamic fee rates and ALM incentive plugins to achieve trading discounts and cross-chain yield aggregation; Jojo Exchange and Uniswap Foundation, on the other hand, leverage ZK verification mechanisms to build a more secure trading incentive system.

At the cross-chain and infrastructure layer, Brevis has expanded from Ethereum to BNB Chain, Linea, Kernel DAO, TAC, and 0G, providing trusted computing and cross-chain verification capabilities for a multi-chain ecosystem. Meanwhile, projects like Trusta AI, Kaito AI, and MetaMask are utilizing the ZK Data Coprocessor to build privacy-preserving point systems, influence scoring, and reward systems, promoting the intelligent development of Web3 data. At the system's underlying layer, Brevis relies on the EigenLayer AVS network to provide re-staking security guarantees and combines the NEBRA Aggregated Proof (UPA) technology to compress multiple ZK proofs into a single submission, significantly reducing on-chain verification costs and latency.

Overall, Brevis has covered the entire life cycle of application scenarios, from long-term incentives, activity rewards, transaction verification to platform services. Its high-frequency verification tasks and reusable circuit templates provide real performance pressure and optimization feedback for Pico/Prism, which is expected to feedback to the L1 zkVM real-time proof system at the engineering and ecological level, forming a two-way flywheel of technology and application.

6. Team Background and Project Financing

Mo Dong | Co-founder, Brevis Network

Dr. Mo Dong is a co-founder of Brevis Network and holds a Ph.D. in Computer Science from the University of Illinois at Urbana-Champaign (UIUC). His research has been published in top international academic conferences, adopted by tech companies like Google, and has received thousands of academic citations. He is an expert in algorithmic game theory and protocol mechanism design, focusing on promoting the combination of Zero-Knowledge Computing (ZK) and decentralized incentive mechanisms, dedicated to building a credible Verifiable Compute Economy. As a venture partner at IOSG Ventures, he has also been focused on early investments in Web3 infrastructure.

The Brevis team was founded by PhDs in cryptography and computer science from UIUC, MIT, and UC Berkeley. Core members have many years of research experience in zero-knowledge proof systems (ZKP) and distributed systems, and have published several peer-reviewed papers. Brevis has received technical recognition from the Ethereum Foundation, and its core module is considered a critical on-chain scalability infrastructure.

Brevis completed a $7.5 million seed round financing in November 2024, led by Polychain Capital and Binance Labs, with participation from IOSG Ventures, Nomad Capital, HashKey, Bankless Ventures, and strategic angel investors from Kyber, Babylon, Uniswap, Arbitrum, and AltLayer.

7. ZKVM and ZK Coprocessor Market Competitor Analysis

Currently, ETHProofs.org, supported by the Ethereum Foundation, has become the core tracking platform for the L1 zkEVM real-time proving (Realtime Proving, RTP) route, used to publicly showcase the performance, security, and mainnet adaptation progress of various zkVMs.

Overall, the competition in the RTP track is focused on four core dimensions:

• Maturity: SP1 is the most mature for production deployment; Pico has leading performance and is close to mainnet standards; RISC Zero is stable but RTP data is not public.
• Performance: The Pico proof has a size of about 990 kB, which is approximately 33% smaller than SP1 (1.48 MB) and at a lower cost;
• Security and Auditing: Both RISC Zero and SP1 have passed independent security audits; Pico is currently undergoing the auditing process;
• Ecosystem Development: Mainstream zkVMs adopt the RISC-V instruction set, and SP1 forms a widely integrated ecosystem relying on the Succinct Rollup SDK; Pico supports Rust for automatic proof generation, rapidly improving SDK completeness.

According to the latest data, the RTP track has formed a “two strong pattern.”

• The first-tier Brevis Pico (including Prism) and Succinct SP1 Hypercube both directly target the P99 ≤ 10s standard set by EF. The former achieves performance and cost breakthroughs with a distributed multi-GPU architecture; the latter maintains engineering maturity and ecological robustness with a monolithic system. Pico represents performance and architectural innovation, while SP1 represents practicality and ecological leadership.
• The second tier RISC Zero, ZisK, and ZKM continue to explore ecological compatibility and lightweight solutions, but have yet to disclose complete RTP metrics (latency, power consumption, CAPEX, security level, proof size, reproducibility). Scroll (Ceno) and Matter Labs (Airbender) are attempting to extend Rollup technology to the L1 verification layer, reflecting the evolutionary trend from L2 scaling to L1 verifiable computation.

By 2025, the zkVM track has formed a technological pattern characterized by RISC-V unification, modular evolution, recursive standardization, and hardware accelerated parallelism. The general verifiable compute layer of zkVM can be divided into three categories:

• Performance-oriented: Brevis Pico, SP1, Jolt, ZisK focus on low latency and real-time proof, enhancing computational throughput through recursive STARK and GPU acceleration.
• Modular and Scalable: OpenVM, Pico, SP1 emphasize modular pluggability, supporting co-processor integration.
• Ecological and General Development Type: RISC Zero, SP1, ZisK focus on SDK and language compatibility, promoting universality.

Comparison of zkVM Competing Projects (as of October 2025)

The current zk-Coprocessor track has formed a pattern represented by Brevis, Axiom, Herodotus, and Lagrange. Among them, Brevis leads with its “ZK Data Co-Processor + General zkVM” integrated architecture, featuring capabilities for historical data reading, programmable computation, and L1 RTP; Axiom focuses on verifiable queries and circuit callbacks; Herodotus specializes in historical state access; Lagrange optimizes cross-chain computing performance with a ZK+Optimistic hybrid architecture. Overall, zk-Coprocessor is becoming a trusted computing interface connecting applications such as DeFi, RWA, AI, and identity in the form of a “verifiable service layer.”

8. Summary: Business Logic, Engineering Implementation, and Potential Risks

Business Logic: Performance-driven and Dual-layer Flywheel Brevis builds a multi-chain trusted computing layer with “general zkVM (Pico/Prism)” and “data coprocessor (zkCoprocessor)”: the former solves any computable verifiable problem, while the latter realizes the business landing of historical and cross-chain data. Its growth logic forms a positive cycle of “performance-ecosystem-cost”: the RTP performance of Pico Prism attracts leading protocol integrations, leading to proof scale growth and a decrease in single transaction costs, forming a continuously strengthening dual-layer flywheel. The competitive advantages mainly lie in three points:

1. Performance reproducibility - has been incorporated into the Ethereum Foundation ETHProofs RTP system;
2. Architectural Barriers - Modular design and multi-GPU parallel implementation for high scalability;
3. Commercial Verification - Scaled implementation has been achieved in incentive distribution, dynamic pricing, and cross-chain verification.

Engineering Implementation: From “Heavy Execution” to “Verification Instead of Execution”

Brevis achieves an average of 6.9 seconds and P99 < 10 seconds in a 45M gas block through the Pico zkVM and Prism parallel framework (64×5090 GPU, <(K CAPEX), with both performance and cost leading the way. The zkCoprocessor module supports historical data reading, circuit generation, and backchain verification, and can flexibly switch between Pure-ZK and Hybrid modes, with overall performance basically aligned with Ethereum's RTP hard standards.

Potential Risks and Key Points of Concern

• Technical and compliance thresholds: Brevis still needs to publicly disclose and undergo third-party verification of hard metrics such as power consumption, security level, proof size, and trusted settings dependency. Long-tail performance optimization remains key, and EIP adjustments may alter performance bottlenecks.
• Competition and substitution risks: Succinct (SP1/Hypercube) still leads in toolchain and ecosystem integration, and the competitiveness of teams like Risc Zero, Axiom, OpenVM, Scroll, and zkSync should not be underestimated.
• Revenue Concentration and Business Structure: Currently, the proof volume is highly concentrated (the top four applications account for about 80%), and it is necessary to reduce dependence by expanding into multiple industries, multiple public chains, and multiple use cases. GPU costs may affect unit gross profit.

Overall, Brevis has built a preliminary moat on both “performance reproducibility” and “business implementability”: Pico/Prism has firmly established itself in the first tier of the L1 RTP race, while the zkCoprocessor has opened up high-frequency, reusable commercialization scenarios. In the future, it is recommended to set achieving the Ethereum Foundation RTP full hard indicators as a phased goal, continuously strengthening the standardization of coprocessor products and ecosystem expansion, while promoting third-party reproducibility, security audits, and cost transparency. By achieving a structural balance between infrastructure and SaaS revenue, a sustainable business growth loop can be formed.