The Brevis ZK Data Coprocessor v2: Unlocking Efficiency and Expressiveness in Intelligent dApps

2025-01-08

[read_meter]

TL;DR:

Our Brevis ZK Data Coprocessor v2 is live on mainnet and introduces a modular architecture combining STARK, Plonk, and Groth16, significantly improving scalability and performance. Many of our partners are already benefiting from this upgrade today. Some of the key benefits include:

Unbounded Data Access: Handle any number and size of historical receipt, transaction, and storage data subproofs in a single request.
Increased Performance (10X): Optimized ZK circuits and parallel proving systems achieve over 10X the throughput and latency improvements on commodity hardware at the same cost as v1.
New Features: Access to full block info, multi-request batching, and enhanced data processing APIs to support new use cases.

How the Brevis ZK Data Coprocessor Works

The Brevis ZK Data Coprocessor enables users to build data-driven and intelligent dApps that can access historical on-chain data points including receipts (i.e. events in a transaction), transactions, and storage slots. Subsequently, dApps can run arbitrary computations using Brevis-generated circuits on these data points.

Developers can build with Brevis without any knowledge of ZK circuits. They can just follow three simple steps to use Brevis in their dApps: 1) use JS APIs to specify what data they want to use, 2) use a highly programmable SDK to express how they want to compute on the data, and 3) write a callback handler function in the smart contract to handle the computation result.

The Brevis ZK Data Coprocessor is highly optimized for dApps to build anything related to on-chain data access and computation with specialized data proof circuits, highly parallel systems, and hardware acceleration. Since our alpha mainnet in September 2023, we have seen many exciting Brevis-supported use cases that focus on next-gen user experiences from forward-looking protocols including PancakeSwap, QuickSwap, Algebra, Usual, Gamma, Kwenta (acquired by Synthetic), Trusta Labs, Mendi Finance, Thena, Jojo Exchange, and Mellow Finance.

The high performance is why many dApps choose Brevis, and we believe it is the core factor driving an inflection point in off-chain verifiable computation. Today, with v2, we deliver ten-fold the performance at the same cost.

What’s New in v2?

Brevis ZK Data Coprocessor v1 relied entirely on pairing-based proving systems (e.g., Plonk/Groth16) implemented with the gnark framework. In contrast, the Brevis ZK Data Coprocessor v2 adopts a hybrid architecture with STARK, Plonk, and Groth16.

This modular and composable v2 architecture also significantly enhances scalability with finer-grained parallelization and an unlimited multi-level subproof recursion pipeline. With this new pipeline, v2 is now able to achieve:

Unbounded Scalability: Process any number of receipt/transaction/storage data subproofs in a single request without upper limits.
Arbitrary Data Sizes: Handle subproofs of any size, such as receipts with numerous event logs or lengthy RLP data.
Improved Efficiency: Achieve over 10X throughput and reduced end-to-end latency using commodity CPU hardware in a scale-out system at the same cost as v1.

Additionally, based on our partners’ feedback and requests, v2 introduces new features like block info export, multi-request batching, and expanded data processing APIs.

With all these upgrades, the Brevis ZK Data Coprocessor is capable of handling use cases involving millions of data points and complex computational logic for our flagship partners.

Benchmarks

The following benchmarks compare v1 and v2 architectures for processing ZK Data Coprocessor requests with varying numbers of receipt data points to generate an EVM-verifiable proof. Tests were conducted on a cluster of 9 AWS c7a.24xlarge machines.

Data Subproofs	End to End Time (v2)	End to End Time (v1)
256 receipt	68s	387s
1024 receipt	127s	1121s
4096 receipt	344s	4057s
8192 receipt	527s	N/A

The v2 architecture delivers more than 10x faster performance on production-like workloads compared to v1, with performance gains becoming increasingly pronounced as the number of data subproofs grows.

What’s Next?

Since its launch, the Brevis ZK Data Coprocessor has enabled groundbreaking applications:

Dynamic fees for Algebra Integral.
Brevis-supported hooks for Uniswap v4 and PancakeSwap v4.
Trustless on-chain reputation by Trusta Labs.
Continuous protocol incentivization for Usual.
Intelligent lending with Mendi Finance.
And many more…

These innovative solutions leverage historical on-chain data to deliver intelligent features responsive to dynamic data input, revolutionizing the way applications interact with the blockchain. This reshapes next-gen user experiences crucial for blockchain mass adoption.”

Looking ahead, Brevis is advancing toward an even more expressive SDK with exciting new initiatives set to be released soon. As Brevis enters v2 and continues evolving, its enhanced performance—improved latency, an unbounded subproof limit, and greater expressiveness—will unlock a new era of possibilities. This transformation will empower developers to build latency-sensitive, data-driven decentralized applications that will expand the boundaries of what’s possible on blockchain, offering users smarter, more efficient, and highly responsive dApps.

About Brevis

Brevis is a verifiable computing platform powered by zero-knowledge proofs, serving as the infinite compute layer for Web3. Applications can offload expensive computations off-chain while proving every result on-chain. The Brevis stack includes Pico zkVM for general-purpose computation, the ZK Data Coprocessor for trustless access to historical blockchain data, Pico Prism for real-time Ethereum block proving (99.6% coverage, 6.9s average), and ProverNet, a decentralized marketplace for ZK proof generation. To date, Brevis has generated hundreds of millions of proofs across 40+ protocols on 6 blockchains.