Qiskit SDK v2.1, which has been released by IBM Quantum, brings major improvements to performance and makes it possible to demonstrate quantum advantage shortly. The release emphasizes the value of ongoing cooperation between the high-performance computing (HPC) and quantum communities, with an emphasis on expanding the support for the C API in the Qiskit SDK. Because C, as opposed to Python, has substantial interchange with other programming languages that are used in HPC, such as C++ and Fortran, this addition is essential because it makes incorporating Qiskit into HPC experiments easier and more flexible.
The key updates in Qiskit SDK v2.1 include
- Extended C API support: Developers can now build quantum circuits and targets directly using a C API, which already demonstrates improved circuit-building time performance. The goal is to eventually support complete workflows in C.
- Transpiler improvements: The preset pass managers are now set up to automatically detect when a target employs a discrete Clifford+T basis.
- Improved synthesis algorithms: These enhancements lead to reduced circuit depth, gate counts, and overall resource consumption.
- Tools for multi-controlled gates: New synthesis algorithms and plugins provide greater flexibility for working with multi-controlled gates.
- BoxOp instruction enhancements: For more expressive timing control, the box instruction now allows stretch durations and custom annotations.
You can also read QEDMA Raises $26 M With IBM To Tackle Quantum Errors
Expanded C API Support for Quantum Circuit Construction: The expanded C API in Qiskit v2.1 allows for the direct construction of quantum circuits in C. This significantly speeds up circuit construction, with a 1000-qubit GHZ state being approximately 10-20 times faster when mixing C and Python compared to a pure Python implementation. The C API currently supports building circuits with standard Qiskit gates and instructions, including measure, delay, reset, barrier, and Unitary Gate. Circuits built in C can be easily passed back into Python using qk_circuit_to_python() for interoperability with Qiskit’s transpiler, visualization tools, and runtime integrations.
Transpiler Support for Clifford+T Targets: The generate_preset_pass_manager() function has been enhanced to automatically determine whether the target basis is composed entirely of Clifford and T gates. In this case, it creates a pass manager that can break down circuits on this discrete basis. The pass manager now employs the BasisTranslator pass and the Solovay-Kitaev decomposition to mimic single-qubit rotations using H, T, and Tdg gates in order to get around the limits in decomposing these rotations. Transpilation may not work or may be less effective if the given Clifford+T basis gate set is not universal. For example, if the S-gate is excluded, it may break down into pairs of T-gates.
Improved Synthesis Algorithms: For multi-controlled operations and decomposition into the Clifford+T basis set, Qiskit SDK v2.1 offers even more performance enhancements in circuit synthesis. Due to its conversion to Rust, the Solovay-Kitaev decomposition transpiler pass has a 100x speedup in runtime, which is a notable increase.
Annotations and Stretch Delays for Boxes: Qiskit v2.1 adds annotations to boxes after introducing the BoxOp instruction in Qiskit SDK v2.0. By allowing users to annotate boxes with specific data for use by later stages of the execution stack or custom transpiler passes, annotations give the circuit constructor a means of communicating intent about how a particular area of the circuit should be handled. Furthermore, boxes may now accommodate durations that make use of stretch variables, which enables more accurate timing control.
You can also read Bias-Tailored Quantum LDPC Codes Boost Quantum Computing
Deprecations and Upgrade Considerations
- Python v3.9 deprecation: Users will need to update to Python v3.10 or later as support for Python v3.9 will be eliminated in Qiskit SDK v2.3.
- Circuit library refactoring: Operations are now classified as either Instruction or Gate objects (specified by action) or QuantumCircuit objects returned by generic Python functions (defined by structure), continuing the reworking that was started in Qiskit v1.3. This includes the official deprecation of various QuantumCircuit subclasses, with n_local() for TwoLocal and QFTGate for QFT as suggested substitutes.
- SymEngine and SymPy are now optional dependencies: With Qiskit SDK v2.1, SymEngine and SymPy are no longer necessary because of a new internal symbolic math library for the Parameter Expression class that is created in Rust. They are still optional for certain applications, including desterilizing older QPY formats or using particular visualization techniques.
Upgrading to Qiskit SDK v1.4 is advised for those still using v1.x. This version provides security support until March 2026 and bugfix support until at least the end of September, giving users enough time to switch to the v2.x series.
You can also read SpeQtral and Thales Alenia Space Connect Quantum Satellites
