Back to blog
Yi Sun
Author
Yi Sun
Published
02 Sep 2025
Category
Product

OpenVM GPU is live on the Axiom Proving API

OpenVM GPU is live on the Axiom Proving API

Today we’re releasing GPU support for OpenVM on the Axiom Proving API. This v1.4.0 release brings substantial performance improvements on both cost and latency by enabling distributed multi-GPU proving and faster execution for OpenVM. OpenVM can now prove Ethereum mainnet blocks in 15s, a 5x+ improvement over the v1.0.0 release. To learn more about the technical and performance updates, check out the OpenVM v1.4.0 announcement.

In conjunction with the OpenVM GPU release, we are also releasing the Axiom Proving API to production. Since the private beta release in May 2025, we’ve added support for projects in the API console, improved backend robustness, and added support for lower latency distributed proving. We are now excited to support developers generating proofs on the API in production.

With this production release, we are continuing to work with developers of ZK applications by providing OpenVM support and custom API deployments. 

  • We’re partnering closely with teams to tailor the most performant ZK solution for their applications with OpenVM, deployed flexibly with the open-source OpenVM GPU prover.
  • We’re onboarding new users for blazing-fast and reliable ZK proving on the Axiom Proving API – reach out to get an invitation to the API console.

If you are interested in using OpenVM or working with us, get in touch, or continue reading to learn more!

What’s New in OpenVM v1.4.0

OpenVM v1.4.0 is a step change in performance from the first production release of OpenVM in March 2025. By introducing end-to-end GPU proving and optimized execution, we’ve cut proving cost and latency by over 5x, achieving latency as low as 15s and cost as low as $0.0003 per transaction for Ethereum mainnet blocks. For detailed performance benchmarks, check out the OpenVM announcement. Latency progress on block 21,000,000 since December is shown in the chart below.

These performance improvements were driven by innovations across the OpenVM stack:

  • End-to-end GPU prover: The new OpenVM GPU prover provides end-to-end acceleration for trace generation and proving and is MIT / Apache 2.0 licensed, meaning developers can deploy it in the most convenient way for their application.
  • Distributed GPU orchestration: OpenVM v1.4.0 supports parallel proof generation across multiple GPU machines for the lowest possible latency. 
  • Optimized execution: The new execution for OpenVM runs at 150 MHz, making it easy for developers to run workloads without generating full proofs.

OpenVM v1.4.0 was audited by Cantina and is live today for production use on the Axiom Proving API. 

Taking the Axiom Proving API to Production

After a four month beta release, we are taking the Axiom Proving API to production with support for OpenVM v1.4.0 on GPU. Developers can use the production API to:

  • Build and Develop OpenVM Programs: Iterate on your ZK application using builds and test proofs on the Axiom Proving API, without the need to set up complex cloud infrastructure during development.
  • Generate Proofs for OpenVM: Generate proofs quickly and efficiently on Axiom’s cloud with optimized GPU orchestration to get maximum performance for your application without getting bogged down by devops and infrastructure overhead.
  • Verify Builds and Deployment: The Axiom Proving API manages cryptographic build artifacts and verification setup to enable you to securely and reproducibly deploy your ZK application. 

If you are an application or infrastructure developer interested in trying out the API to make your ZK more easily and efficiently, get in touch or check out the docs to learn more. 

Work with Us

We are working with teams to achieve the best possible ZK performance by customizing OpenVM and the Axiom Proving API to integrate deeply with their application. We believe that successful ZK applications will need tailored ZK solutions across the VM, proof generation, and deployment layers, and this has informed the open and customizable design philosophy underlying OpenVM. If you are interested in working closely with our team to build a ZK solution for your needs, reach out here!

In the meantime, try out OpenVM using the book, specs, and Github, and join the developer chat on Telegram for questions and technical discussion. At Axiom, we’re excited to continue pushing the frontier of what’s possible in ZK with OpenVM.