zkvm-benchmarks

Unified benchmarking framework for Zero-Knowledge Virtual Machines (zkVMs). Measures execution time, proof metrics, and resource usage across implementations using standardized guest programs.

Support zkVMs

zkVM	Version	Status	Programs	Notes
SP1	v5.2.2	🟢 Ready	All	Production-ready SDK
Risc0	v3.0.3	🟢 Ready	All	Production-ready SDK
Jolt	v0.3.0-alpha	🟢 Ready	All	Full SDK Support
Nexus	v0.3.4	🟢 Ready	All	Full SDK Support
Pico	v1.1.6	🟢 Ready	All	Full SDK Support
Powdr	v0.1.0	🟢 Ready	All	Full SDK Support
ZKM	v0.3.0	🟢 Ready	All	Network/Local Proving
zkWasm	v0.1.0	🟢 Ready	All	Delphinus CLI
Valida	v1.0.0	🟢 Ready	All	CLI Integration
SnarkVM	v0.16.0	🟢 Ready	All	Aleo SDK
Miden	v0.11.0	🟢 Ready	Fib, Sum, Fac	MASM Support
Cairo	v2.8.5	🟢 Ready	All	Stwo/Cairo SDK
Airbender	v0.5.0	🟢 Ready	All	Nightly Rust
Zisk	v0.10.0	🔴 Linux	All	No macOS support
Ceno	v0.1.0	🟡 Ref	All	SDK Private/Simulated
Lean	v0.1.0	🟡 Ref	All	SDK Private/Simulated
Cairo-M	v0.1.0	🟡 Ref	All	Simulated Proofs
Novanet	v0.1.0	🟡 Ref	All	WASM Runner
o1vm	v0.1.0	🟡 Ref	All	Framework Only
OpenVM	v1.4.0	⚪ Off	All	Disabled by default

Legend: 🟢 Ready (Full Integration) | 🟡 Ref (Reference/Simulated) | 🔴 Linux (OS Dependent) | ⚪ Off (Disabled)

Quick Start

Notes: Before running benchmarks, make sure your environment can run individual zkVMs natively (e.g., you can run SP1 or Risc0 programs without Docker). This ensures that the benchmarks will run smoothly and generate valid results.

• Haredware Requirements

  • CPU: 16+ cores
  • RAM: 32GB+ (for heavy zkVMs like SP1)
  • Storage: 100GB+ free space (for SDKs, toolchains, and benchmark results)
  • OS: macOS/Linux (recommended: macos)

• Prerequisites

  • Rust toolchain
  • golang toolchain
  • Docker
  • Make,curl,git etc.

• Install zkVM SDKs/Toolchains

  Run benchmarks natively needs installed zkVM toolchains/SDKs. Install them using the provided scripts.

  cd scripts/sdk_installers
  bash install_xxx.sdk.sh

  If you meet any problems, we recommend to reference official docs of related zkVM to set up the environment and run their example programs, which can help you understand the underlying mechanics and debug potential issues during benchmarking.

• Run

  Run benchmarks natively using the provided Makefile.

  cd zkvm_benchmark_utils
  
  # 1. Run specific zkVM (e.g., SP1)
  make bench_sp1
  
  # 2. Run fast benchmark set (SP1, Risc0, Jolt)
  make bench_fast
  
  # 3. Run all enabled zkVMs
  make bench_all

• Result

  You can find the results in the zkvm_benchmark_utils/benchmark-results/reports directory. Each run will generate a timestamped subdirectory containing raw logs, parsed metrics, and generated reports. And the final reporter(csv) can be found at zkvm_benchmark_utils/benchmark-results/reports, organized by zkVM, program and input.

Benchmark Utils

For details on metrics, configuration, and advanced usage, see: 👉 zkvm_benchmark_utils/README.md

Contributing

Welcome contributions from the community to expand zkVM support, add new benchmark programs, and improve the framework.

Acknowledgements

This project aims to provide a unified benchmarking framework for measuring and comparing the performance of different zkVM implementations and their example programs on custom metrics such as execution time, proof metrics, and etc. During the development, The following resources and projects have been invaluable for reference and inspiration:

• https://github.com/blocksense-network/zkVMs-benchmarks
• https://github.com/kkrt-labs/zkvm-benchmarks

We acknowledge the efforts of the zkVM community in developing these innovative technologies and hope that this framework can contribute to their evaluation and improvement.