Proof-based general-purpose computing with Magnetar
A Microsoft zkVM proves ordinary Rust/Go/C programs 10-39x faster than prior systems, using only standard hash functions, no…
Microsoft Research introduced Magnetar, a zero-knowledge virtual machine that proves the execution of programs compiled to RISC-V (written in Rust, Go, or C) without requiring developers to hand-write cryptographic circuits. The company reports it is 10 to 39 times faster on average than the fastest prior comparable zkVMs across 15 benchmarks, proves millions of instructions per second on a laptop, and relies only on standard hash functions rather than elliptic curves or novel cryptographic assumptions, making it post-quantum-resistant by design.
Why it matters
Zero-knowledge proofs let a result be trusted because of what accompanies it (a cheaply verifiable proof) rather than who produced it, which matters for verifiable AI agent actions, confidential compliance checks, and blockchain execution. Magnetar's claimed speedup and reliance on well-studied cryptography (versus newer, less-scrutinized ZK-friendly primitives) would lower both the performance and the long-term security risk of adopting general-purpose verifiable computing, if the benchmarks hold up under independent scrutiny.
What is verified
Magnetar is a zkVM that proves programs compiled to RISC-V and written in Rust, Go, or C, without requiring developers to write cryptographic circuits. [1]
Across 15 benchmark programs, Magnetar is reported to be 10 to 39 times faster on average than the fastest existing zkVMs on the same machine, and over 300 times faster than two others on a token-transfer benchmark. [1]
Magnetar's security rests only on standard hash functions such as SHA-256 or SHA-3, requiring no elliptic curves, lattices, trusted setup, or novel cryptographic assumptions. [1]
Magnetar proves 7.2 to 8.7 million processor instructions per second on a laptop, with proofs verifying in about 12 milliseconds. [1]
What is not
- This is a Microsoft Research blog post, not an independently peer-reviewed paper or third-party benchmark; no external replication is cited.
- The comparison baseline systems and exact benchmark methodology (hardware specs, workload selection criteria) are only partially described.
- No release date, availability, licensing, or open-source status for Magnetar is given, so it is unclear whether or when developers can actually use it.
- No concrete deployment case studies are given beyond illustrative examples (AI agent policy compliance, financial transaction proofs, EVM interpreter benchmark).
Sources
- 1Microsoft Researchmicrosoft.com · Primary source
Original source published 15 October 2026 at 18:00.
An AI-assisted briefing by Coding Partners. We link the evidence behind the reporting so you can explore the original work and assess the claims. How CP Signal works.