Quantum Computing Report

IBM Releases Qiskit v2.3 with Expanded C API and Fault-Tolerant Primitives

IBM has released Qiskit SDK v2.3, prioritizing deeper integration with High-Performance Computing (HPC) environments and the development of fault-tolerant compilation pipelines. A central feature of this release is the expansion of the C API, introducing the QkDag object and an updated QkTarget model. These tools allow developers to write and execute custom transpiler passes directly in C, enabling granular circuit optimization without requiring a full compiler pipeline rebuild. This facilitates the integration of Qiskit into existing C-based HPC software stacks and custom hardware workflows.

The release includes Rust-driven performance enhancements to circuit-to-hardware layout selection. Specifically, updates to VF2Layout and VF2PostLayout improve the speed and scalability of mapping quantum circuits to physical hardware topologies. These optimizations are designed to reduce compilation overhead and improve gate fidelity by selecting more efficient qubit mappings. Additionally, the transition of the ControlFlowOp to Rust is now complete, finalizing the refactor of Qiskit’s internal data model and positioning the SDK for future speed gains in complex, dynamic circuit management.

Qiskit v2.3 introduces primitives essential for large-scale, fault-tolerant architectures. The new PauliProductMeasurement instruction enables joint projective measurements across multiple qubits, a prerequisite for Pauli-based computation (PBC) and error-corrected protocols. Furthermore, the transpiler now supports the Ross-Selinger (gridsynth) algorithm for efficient RZ-rotation approximation in Clifford+T basis sets. The release also unifies gate cancellation logic into the CommutativeOptimization pass, which leverages commutativity to simplify circuits and minimize costly operations like T-gates in early fault-tolerant instruction sets.

System requirements have been updated, with Python 3.10 or higher now required following the end-of-life for Python 3.9. Platform support tiers have also shifted, with macOS x86-64 (Intel) support downgraded from Tier 1 to Tier 2. While pre-compiled wheels remain available for Intel-based Macs, testing is now performed only at the time of release rather than at every code change. These shifts reflect the project’s transition toward Rust-native performance and modern software dependencies for quantum development.

Read the official technical release summary from IBM here and view the feature overview on LinkedIn here.

January 22, 2026

Exit mobile version