HomeTech NewsQuantumIBM and AMD Release Pelagos for Quantum Decoder Tests

IBM and AMD Release Pelagos for Quantum Decoder Tests

IBM and AMD have released Pelagos, an open-source framework for evaluating quantum error-correction decoders at scale. The companies described the 16 November release, in an IBM newsroom post by Blake Johnson, distinguished engineer for quantum error-correction infrastructure at IBM Quantum, and Robert Davis, as a GPU-accelerated research tool that helps researchers run large decoder-evaluation campaigns. It is a software framework, not a new quantum processor and not a claim that a fault-tolerant machine is now in operation.

IBM said the open-source release is being made at SC26, as a further step after the quantum error-correction co-design platform that Oak Ridge National Laboratory, IBM and AMD announced earlier this month. That earlier platform is already a TSN subject. The code is on GitHub.

What Pelagos does, and what the companies say it is for

Quantum computers are prone to errors. Error-correcting codes try to protect quantum information by checking continuously for signs that something has gone wrong. Those checks produce data known as syndromes, which point to where an error may have happened without revealing the underlying quantum information itself. A decoder then analyses the syndromes and decides what correction is most likely. IBM’s newsroom post explains that proving a decoder works becomes more computationally demanding as the codes get larger, and that researchers often need to evaluate billions of sampled error patterns under realistic conditions. Those studies, it said, can require large simulation campaigns and substantial classical computing resources.

Pelagos, named after the Greek word for “open sea,” is the companies’ answer to that evaluation problem. IBM said the framework was built to make the work “faster, more scalable, and more accessible to the broader research community,” and that it combines GPU acceleration with a developer-friendly software architecture. It is built using the Triton programming framework and, IBM said, supports both AMD and NVIDIA GPU environments. That is a statement about the software’s intended reach. It is not a ranking of those GPUs, and it is not a hardware launch.

An AMD technical article published on the same day describes the same release as an open-source tool for accelerating large-scale quantum error-correction decoder simulations. AMD said the team used Triton to build a high-throughput implementation of Relay-BP, a decoder designed for quantum low-density parity check, or qLDPC, codes, and that the project includes a self-tuning GPU decoder with multi-GPU support, integration with the Stim simulator, and step-by-step tutorials. AMD spells the name Pêlagos in that article; IBM, the GitHub repository and the accompanying papers use Pelagos, which is the spelling used in this report except when quoting AMD.

Alongside the release, IBM and AMD are publishing a paper on how the framework scales across multiple GPUs, including deployments on AMD-powered high-performance computing systems, and an accompanying survey paper on decoding systems, classical infrastructure and hardware-software co-design in fault-tolerant quantum computing. IBM’s key takeaways say the new research “demonstrates efficient multi-GPU scaling for Pelagos on AMD Instinct accelerators and ORNL HPC systems.” This report does not attach a speed-up figure to that claim.

How the release sits beside IBM’s other quantum work

The closest earlier TSN account is the ORNL, IBM and AMD quantum error-correction co-design platform, announced earlier this month. IBM’s newsroom post presents Pelagos as another step in that collaboration: a shared software tool for testing decoders, rather than a new processor or a new laboratory installation. Readers who want the hardware-and-platform file should start there.

A separate, newer TSN story covers a planned quantum codesign platform from NVIDIA, IBM and Fermilab. That post is about a codesign effort with a different set of partners and a different brief. It is useful context for IBM’s wider classical-and-quantum work, but it is not the same announcement as Pelagos and should not be read as a companion hardware milestone for this software release.

IBM’s own hardware roadmap remains a third, separate thread. TSN has already reported that IBM plans Nighthawk r3 and a 1,040-qubit linked system. That is a plan for machines. Pelagos is a tool for studying decoders on classical GPUs. The two should not be collapsed into a single “IBM quantum advance.”

A decoder-testing framework is not a hardware milestone

The firm facts are narrow. IBM and AMD have released an open-source, GPU-accelerated framework for evaluating quantum error-correction decoders, described it at SC26 as a further step after the Oak Ridge co-design platform, placed the code on GitHub, and published two accompanying papers, one of them on multi-GPU scaling. IBM says the tool supports AMD and NVIDIA GPUs and was built with Triton. AMD, in its own article, describes a Relay-BP implementation, Stim integration and tutorials. None of that is a new quantum chip, a new dilution refrigerator, or a demonstration that a fault-tolerant quantum computer is running.

It is easy to over-read a release of this kind, especially when it is timed to a supercomputing conference and sits near other IBM quantum announcements. A framework that helps researchers test decoders at scale is useful only if those tests are actually run, compared and published. The survey paper may be a helpful map of the field; it is not itself a result on a particular code or a particular machine. The Oak Ridge platform and the NVIDIA, IBM and Fermilab codesign plan are neighbouring stories, not evidence that Pelagos has been deployed as a production decoder on a live system.

IBM’s key takeaways emphasise “efficient multi-GPU scaling.” That is the companies’ characterisation of the accompanying research, not a TSN measurement, and this report does not convert it into a headline speed-up. The next official facts to watch are ordinary ones: whether independent groups adopt the GitHub code, whether later papers report decoder-evaluation campaigns that used it, and whether the Oak Ridge platform or another laboratory names Pelagos as part of a concrete workflow. Until then, the news is that IBM and AMD have released a research tool for testing decoders, not that they have crossed a hardware threshold.

Sources

Related stories

Follow @tsnmediahub on X for breaking AI news as it happens.

Follow on X

Share this story

Latest stories

  1. TensorWave and OpenAI Agree a 457MW AMD MI455X Deal
    AI Chips & Compute

More in this category

Latest stories

  1. TensorWave and OpenAI Agree a 457MW AMD MI455X Deal
    AI Chips & Compute

Explore the TSN Atlas: Every topic we cover, mapped. Click a bubble to dive in.