snark.fast is a competitive benchmarking platform that measures and ranks post-quantum proof-generation performance for Ethereum, specifically tracking compression throughput on x86 hardware. The platform hosts a public leaderboard where solvers submit optimized implementations, with the current top record achieving 1,651,267 compressions per second—a 312% improvement over the launch baseline. Each submission is verified through a standardized process involving 100 fresh worker runs and a pinned trusted verifier.
Funding
Funding not disclosed
Founders
Product
Problem
Post-quantum cryptographic schemes for Ethereum face significant performance bottlenecks, particularly in proof generation and compression throughput on standard x86 hardware. Without standardized, verifiable benchmarks, developers struggle to compare optimization approaches, slowing the adoption of quantum-resistant solutions.
Solution
snark.fast operates a public benchmarking platform that measures post-quantum proof compression performance for Ethereum on x86 architecture. The platform runs a rigorous evaluation protocol: each submission executes 262,144 compression operations across 100 fresh worker runs, with the median time recorded and every proof validated by a pinned trusted verifier. Results are displayed on a live leaderboard that tracks throughput improvements, with the current top submission achieving 1,651,267 compressions per second—a 312.23% increase over the launch baseline. The platform also maintains per-solver contribution ledgers and run notes, enabling transparent tracking of optimization progress over time.
Target Audience
Primary users are cryptographic engineers, zero-knowledge proof developers, and blockchain infrastructure teams working on post-quantum optimization for Ethereum and related systems.
Features
- Standardized benchmark protocol using 262,144 compression operations per trial across 100 fresh worker runs
- Pinned trusted verifier that checks every proof before results are accepted
- Public leaderboard ranking submissions by throughput, with 185 promoted submissions from 21 solvers
- Per-solver contribution ledgers and detailed run notes for each submission
- Support for multiple AI-assisted solver approaches, including Fable, GPT, DeepSeek, and Opus model variants
- Proof size tracking, with current submissions averaging 429 KiB per proof