What happened
NVIDIA introduced TCBP-JOBS, a decoding algorithm for quantum error correction (QEC). The company says it speeds up research workloads by up to 13,000 times compared with leading existing methods while maintaining high accuracy. The algorithm is available through the CUDA-Q QEC 0.9.0 library.
Why it matters
Qubits are highly sensitive to noise. Error correction protects information by spreading it across multiple physical qubits, and a classical decoder then analyzes measurement data to identify likely errors. NVIDIA says faster decoding lets researchers test different error-correction codes and decoder configurations far more quickly, cutting the time to study new designs from 12 weeks to 45 minutes.
Work with IBM and Fermilab
NVIDIA, IBM and the US Department of Energy's Fermilab plan to build a modular quantum-classical codesign platform. It will colocate NVIDIA DGX B200 systems with IBM quantum computers, including those using its Nighthawk processor. Fermilab will contribute high-energy and nuclear physics expertise.
Fitting into existing workflows
NVIDIA is extending the open-source Slurm workload manager with a quantum interface, while IBM contributes its Quantum Resource Management Interface. The goal is to let researchers schedule quantum processors with familiar commands. IBM, RIKEN and NVIDIA also reported a 79-fold accuracy improvement over earlier runs in a study of an iron-sulfur cluster found in proteins.



