IBM pushes Qiskit further beyond Python with SDK v2.6 release

IBM has released Qiskit SDK v2.6, a software update centered on the project’s low-level C interface rather than its better-known Python layer. According to IBM’s October 22 release summary, the update adds a pass-manager framework to the C API, brings QPY circuit serialization into compiled-language workflows, and lets developers define custom quantum operations directly from C. Taken together, the changes continue IBM’s effort to make more of Qiskit usable without a Python runtime.

What was announced

The v2.6 release extends Qiskit’s C API, which IBM describes as a low-level interface to Qiskit’s Rust-based core data model for use from compiled languages. IBM says the interface is intended both for standalone use without Python and for building higher-performance Python extensions that interact directly with Qiskit’s Rust objects.

In Qiskit v2.6, IBM added a generic pass-manager framework to that C API. The new components include QkPassManager, QkPassVtable, and QkIrHandle, which together allow developers to create compiler passes and chain them into type-checked pipelines across multiple intermediate representations.

Get the DesignQuantum newsletter

Free quantum computing news and explainers, with every claim linked to its source. Choose daily, weekly or monthly.

Website

Email address FrequencyDailyWeeklyMonthly LanguageEnglishEspañol简体中文हिन्दीالعربيةPortuguês (Brasil)FrançaisDeutsch日本語Русский한국어Bahasa IndonesiaTürkçeItalianoTiếng ViệtPolskiNederlandsไทยفارسیУкраїнськаবাংলাاردوKiswahiliBahasa Melayuעברית Subscribe free

Free. Unsubscribe with one click at any time. Privacy policy

The release also exposes QPY serialization functions in the C API, enabling programs written in languages such as C, C++ and Julia to write and read Qiskit’s binary circuit format without round-tripping through Python. In addition, developers can now define custom operations through a C vtable interface, including hardware-native operations, according to IBM.

The technical details

IBM says the new pass-manager framework can validate a compilation pipeline before execution by checking the declared input and output representations for each pass. Built-in representations include QkCircuit and QkDag, while user-defined IRs can also be described through QkIrHandle.

That matters because Qiskit’s compiler work is increasingly organized around lowering programs through several representations, a pattern that is relevant both to current transpilation flows and to future fault-tolerant toolchains.

On serialization, Qiskit 2.6 makes QPY format v18 the default. IBM says this version reduces the size of serialized circuits by using a more compact encoding for quantum and classical registers. Circuits produced through the C API remain compatible with Python’s qiskit.qpy module, according to the company.

The release keeps the minimum supported Python runtime at 3.10, while builds from source now require Rust 1.89. IBM’s roadmap page also notes that Qiskit 2.6.0 was released on October 16, 2026 and that bug fixes will be delivered through patch releases or the next minor release.

Why it matters

For most Qiskit users, Python remains the main entry point. But IBM is clearly building toward a different deployment model as well: compiled-language access to the same core compiler and execution stack.

That shift is relevant for high-performance computing environments, where Python can be a packaging, runtime or integration constraint. IBM’s documentation says the C API is designed to provide a defined ABI to the internal Rust data model, and the company’s blog frames v2.6 as another step toward using Qiskit in environments where Python is not installed.

It also signals a broader architectural transition. The Qiskit roadmap shows IBM working toward formal ABI and API stability for the C interface in v2.7, along with a new middle-end IR for fault-tolerant compilation. For v2.8, the roadmap points to BackendV3, described there as a Rust-native backend interface exposed to Python and C.

What to watch

The immediate question is stability. IBM’s current C API documentation says the interface should still be considered experimental and may change between minor releases. That caveat is important for software teams considering deep integration.

The next milestone, then, is Qiskit 2.7. According to the GitHub roadmap, that version is expected to define stability guarantees for the C API and expand compiler functionality further. If IBM delivers that, Qiskit’s non-Python path will start to look less like an internal plumbing project and more like a production interface for compiled quantum software stacks.