See, customize, and extend your error mitigation

Error mitigation that once ran out of sight on IBM servers now runs on your client—see it, customize it, extend it. The directed execution model is the foundation of the IBM Quantum error correction stack, powering many of the tools described in this edition of our quarterly update. Your circuits run exactly the way you direct them.

Explore the docsRead the blog

Run your circuits on Nighthawk r2

New!

The platform for error correction research—now with 25x Heron’s throughput and support for 7,500+ gate circuits. Run your circuits with all paid plan tiers.

Tutorial

Cut PEC sampling overhead

Probabilistic error cancellation (PEC) returns unbiased results, but at a steep sampling cost. The directed execution model gives you endless ways to trim it via the Executor primitive. Explore two of those methods with the tutorials below: use shaded lightcones (SLC) to leave low-impact errors unmitigated and cut overhead ~3.4x, or pair error detection with PEC via postselected noise channels.

Shaded lightcones tutorial(opens in a new tab)PEC with logical noise models(opens in a new tab)

Tool

Build custom mitigation pipelines

Assemble your own error mitigation pipeline with Qiskit Mitigation, an open-source package including PEC, TREX, and ZNE with PEA and gate-folding amplification. Filter non-Markovian noise with bit-flip checks or swap in a custom extrapolator for a ZNE workflow—without rewriting the protocol. Pair it with Qiskit Noise Learning for per-region Pauli-Lindblad noise characterization. Work at the level you want.

Qiskit Mitigation docs(opens in a new tab)Qiskit Noise Learning docs(opens in a new tab)

Docs

Run primitives locally

Update your provider for the new client-side Sampler and Estimator. Local mode now applies the same error mitigation you'd get on hardware—set a resilience level, compare strategies, and prototype before spending QPU time.

Local testing mode docs(opens in a new tab)

GitHub

Migrate old code with AI

Haven't touched your Qiskit code in a while? Point your AI coding assistant at our first official Qiskit skill and migrate your qiskit-ibm-runtime code to the new client-side primitives—breaking changes flagged for you.

Get the skill on GitHub(opens in a new tab)

Additional updates

Watch the Qiskit Global Summer School 2026 lectures

Missed the Qiskit Global Summer School? Watch the full 2026 lecture series on YouTube—from qubits, superposition, and entanglement to noise, hybrid quantum-HPC workflows, and algorithms for simulation and optimization. Follow hands-on Qiskit examples on real hardware.

Watch on YouTube(opens in a new tab)

Take on QOBLIB's optimization decathlon

Explore QOBLIB, an open-source "intractable decathlon" of 10 challenging optimization problem classes and 1,200+ instances. Solve a market-split benchmark with Kipu Quantum's Iskay optimizer on real hardware, then submit your own best-known results to the live leaderboard.

Start the tutorial(opens in a new tab)Explore the library(opens in a new tab)

Run Qiskit in Fortran, C++, and Julia

Skip the Python layer. Qiskit's C API now reaches Fortran, C++, and Julia, so you can build and run circuits natively in the language your HPC code already speaks. Start with the following tutorial, which provides a gentle on-ramp to running a 60-qubit dynamics simulation, end to end in Julia.

Try the tutorial(opens in a new tab)View the repo(opens in a new tab)