The short version

Start with
Qiskit, unless quantum machine learning is your main goal.
Free hardware
Qiskit. IBM's Open Plan includes runtime on real devices.
Differentiable
PennyLane. Circuits train like PyTorch layers.
Job postings
Qiskit appears far more often than PennyLane.
Both?
Yes. Most practitioners end up using both.

01 The short answer

Most learners should start with Qiskit. Here is when each framework is the better choice:

Learn Qiskit if you want:

  • Access to IBM quantum hardware (free, real devices)
  • The largest beginner community and the most tutorials
  • General-purpose quantum programming skills
  • The most answers on Stack Overflow and Slack
  • The highest probability of finding job postings that list your framework

Learn PennyLane if you want:

  • Quantum machine learning (QML) and differentiable programming
  • PyTorch, JAX, or TensorFlow integration with quantum circuits
  • Hardware-agnostic code that runs on any backend
  • Clean, functional API for variational circuits
  • Research focus on QML or variational algorithms

02 Side-by-side comparison

Feature Qiskit PennyLane
Developer IBM Xanadu
Primary use case General quantum programming, IBM hardware Quantum ML, differentiable circuits
Hardware access IBM quantum devices (free), simulators Multiple backends via plugins (Qiskit, Cirq, etc.)
QML support Basic, not the focus First-class: PyTorch, JAX, TensorFlow integration
Learning resources Extensive: textbook, IBM Learning, hundreds of courses Good: PennyLane Codebook, demos, growing course selection
Community size Very large (500k+ IBM Quantum users) Smaller, concentrated in QML research
Job market More job postings explicitly mention Qiskit Fewer postings, stronger in QML research roles
Best for Beginners, general quantum developers, IBM hardware users QML researchers, differentiable programming, multi-backend

03 The same circuit in both frameworks

The clearest way to feel the difference is to write the same program in each. Below is a Bell state, the simplest entangled two-qubit circuit, built and sampled 1,000 times in Qiskit and in PennyLane. Both produce roughly half 00 and half 11 outcomes. Notice how Qiskit separates the circuit from the primitive that runs it, while PennyLane wraps the circuit in a decorated function that behaves like any other Python callable.

Qiskit (1.x)

from qiskit import QuantumCircuit
from qiskit.primitives import StatevectorSampler

qc = QuantumCircuit(2)
qc.h(0)
qc.cx(0, 1)
qc.measure_all()

sampler = StatevectorSampler()
result = sampler.run([qc], shots=1000).result()
counts = result[0].data.meas.get_counts()
print(counts)  # ~ {'00': 500, '11': 500}

PennyLane (0.4x)

import pennylane as qml

dev = qml.device("default.qubit", wires=2, shots=1000)

@qml.qnode(dev)
def bell():
    qml.Hadamard(wires=0)
    qml.CNOT(wires=[0, 1])
    return qml.counts()

print(bell())  # ~ {'00': 500, '11': 500}

For circuits this small the two look similar. The gap widens with quantum machine learning: in PennyLane, adding a trainable parameter and a gradient-based optimizer is a few lines because the qnode is differentiable end to end, whereas in Qiskit you wire the circuit into a separate optimization loop. Work through both hands-on with our free Qiskit Hello World and PennyLane Hello World tutorials.

04 Decision tree: which should you pick?

Answer these in order. Stop at the first match.

  1. Is your main goal quantum machine learning, or training variational circuits with PyTorch, JAX, or TensorFlow? Yes, go with PennyLane.
  2. Are you brand new to quantum computing and want the smoothest on-ramp with the most tutorials and community help? Yes, go with Qiskit.
  3. Do you need free access to real quantum hardware out of the box? Yes, go with Qiskit (IBM's free Open Plan includes runtime on real devices).
  4. Do you need to run the same code across many hardware vendors without rewriting it? Yes, lean PennyLane (backend-agnostic by design).
  5. Still unsure? Start with Qiskit for the fundamentals, then add PennyLane when you reach variational algorithms or QML. Most practitioners end up using both.

05 Qiskit: strengths and weaknesses

Qiskit is the most widely used quantum computing framework by a large margin. Built and maintained by IBM, it is the official SDK for accessing IBM's real quantum computers.

Strengths

  • Free real hardware access. IBM Quantum's free Open Plan includes 10 minutes of runtime per month on real superconducting devices, including 156-qubit Heron processors. No other framework gives you this out of the box.
  • Largest community. More tutorials, more Stack Overflow answers, more Slack channels, and more courses are built around Qiskit than any other framework.
  • Qiskit Runtime for production. IBM's Qiskit Runtime platform provides production-grade primitives (Sampler, Estimator) and error mitigation tools for running on real hardware.
  • Native error mitigation. Built-in tools for zero-noise extrapolation, measurement error mitigation, and probabilistic error cancellation.

Weaknesses

  • IBM-centric. Qiskit is optimized for IBM hardware and IBM's ecosystem. Running on other hardware vendors requires additional plugins and is less seamless than PennyLane's backend-agnostic design.
  • Less elegant for QML. Integrating Qiskit circuits with PyTorch or JAX for gradient-based optimization is possible but requires more boilerplate than PennyLane's native differentiable design.
  • Qiskit 1.0 API changes. The transition to Qiskit 1.0 in 2024 introduced significant API changes. Many older tutorials and StackOverflow answers use deprecated patterns that no longer work, check publication dates carefully.

06 PennyLane: strengths and weaknesses

PennyLane, built by Xanadu, was designed from the ground up around differentiable programming, treating quantum circuits as machine learning layers that can be trained with gradient descent.

Strengths

  • Differentiable programming. PennyLane circuits are natively differentiable with autograd, PyTorch, JAX, and TensorFlow. This makes writing quantum gradient descent loops as natural as writing classical neural network training loops.
  • Best for QML research. If you are working on quantum neural networks, variational classifiers, or quantum generative models, PennyLane is the standard framework in the research community.
  • Hardware-agnostic. Write once, run on any backend: IBM, Google Cirq, Amazon Braket, Rigetti, or PennyLane's own simulators. The backend-agnostic design means your code is not locked to one vendor.
  • Cleaner variational circuit API. PennyLane's functional design makes writing parameterized circuits and hybrid quantum-classical optimizers more concise and readable than equivalent Qiskit code.

Weaknesses

  • Smaller community. Fewer tutorials, smaller Stack Overflow presence, and a less active general-purpose community than Qiskit.
  • Fewer job postings. Explicit mentions of PennyLane in job descriptions are much rarer than Qiskit. Most quantum software roles want general quantum programming skills or Qiskit specifically.
  • Less hardware access out of the box. PennyLane does not give you free access to real quantum hardware directly, you need IBM Quantum credentials or another cloud provider account, accessed via a plugin.

07 Where do Cirq, Braket, and other frameworks fit?

Qiskit and PennyLane are the two most common starting points, but they are not the only frameworks, and picking one does not lock you out of the others. Here is how the rest of the landscape relates to this decision.

  • Cirq (Google). A general-purpose circuit framework like Qiskit, built for low-level control and Google's hardware. If you are choosing a general SDK and not tied to IBM, Cirq is the main alternative to Qiskit. See the Google quantum platform and our Cirq Hello World tutorial.
  • Amazon Braket. Not really a competing framework but a cloud service that gives one paid API to many hardware vendors (IonQ, IQM, Rigetti, QuEra). You often write circuits in PennyLane or Braket's SDK and submit them through Braket. See the AWS quantum platform.
  • Cirq vs PennyLane, Braket vs Qiskit, and so on. Because PennyLane is backend-agnostic and Qiskit, Cirq, and Braket all expose hardware, these frameworks interoperate more than they compete. PennyLane can run circuits on IBM, Google, and Braket backends through plugins.

For a full neutral breakdown across every major framework and platform, see our framework and platform comparison.

08 Can you learn both?

Yes, and many practitioners do. The core concepts of quantum circuits, qubits, gates, measurement, superposition, entanglement, are the same across frameworks. Once you understand them in one framework, picking up the other takes days, not months.

Once you understand them in one framework, picking up the other takes days, not months.

The recommended path for most learners: start with Qiskit for the foundations (2-4 months), then add PennyLane if you find yourself working on variational algorithms, QML, or multi-backend deployments. If you know you want to focus on QML from the start, you can go PennyLane-first, but expect a steeper early ramp due to fewer beginner resources.

Browse all Qiskit resources or all PennyLane resources for a comprehensive view of what is available on each platform.

09 Top Qiskit courses

Highest-rated courses for learning Qiskit and IBM quantum programming.

  • IBM Quantum

    Quantum Optimization with IBM Quantum (openHPI)

    Julien Gacon, Dr. Daniel J. Egger, Dr. Stefan Woerner, Lucia Cuervo Valor (IBM Quantum)

    Free openHPI course on quantum approaches to combinatorial optimization, developed with IBM Quantum. Covers QAOA, VQE, and the quadratic unconstrained binary optimization (QUBO) framework with Qiskit implementation.

    2 weeks, 3 to 5 hours per week advanced

10 Top PennyLane courses

Highest-rated courses for learning PennyLane and quantum machine learning.

  • PennyLane

    PennyLane Codebook: Interactive Quantum Computing Lessons

    Xanadu / PennyLane Team

    Interactive browser-based coding exercises teaching quantum computing with PennyLane from scratch, spanning 15 modules from qubits and gates through Grover's and Shor's algorithms, error correction, and variational quantum algorithms.

    Self-paced beginner

  • PennyLane

    PennyLane Quantum Machine Learning Demos

    Xanadu / Community

    Xanadu's collection of hundreds of interactive PennyLane tutorials, including 50+ quantum machine learning demos covering VQE, QAOA, quantum kernels, quantum neural networks, and more.

    Self-paced intermediate

11 Frequently asked questions

Should I learn Qiskit or PennyLane first?
Start with Qiskit if you are new to quantum computing. Qiskit has more beginner resources, a larger community, more Stack Overflow answers, and more tutorials that cover the quantum computing fundamentals (gates, circuits, measurement) in depth. The IBM Learning platform, Qiskit Textbook, and dozens of courses are built around Qiskit. Once you are comfortable with quantum circuit basics and want to explore quantum machine learning or variational algorithms with PyTorch or JAX integration, add PennyLane. Many practitioners use both.
Can PennyLane use IBM quantum hardware?
Yes. PennyLane is hardware-agnostic by design and includes a Qiskit plugin (pennylane-qiskit) that lets you run PennyLane circuits on IBM's real quantum devices and simulators. You can write your circuit in PennyLane and execute it on IBM hardware without switching frameworks. This is one of PennyLane's key strengths: you can develop and test with one backend and run on another. The plugin is actively maintained by Xanadu.
Is Qiskit still the most popular quantum framework?
Yes, by a wide margin in terms of downloads, community size, and course availability. Qiskit has over 500,000 registered users on IBM Quantum, thousands of tutorials, and the largest community on Slack and Stack Overflow. PennyLane has grown rapidly in the QML research community but remains smaller overall. For job postings that mention a specific framework, Qiskit appears far more often than PennyLane or any other framework.
Which framework has better job prospects?
Qiskit skills are currently more frequently listed in job postings for quantum computing roles. Most quantum software engineering positions that specify a framework require Qiskit or general quantum circuit knowledge. PennyLane is more valuable for quantum machine learning research roles, particularly in academic and startup settings focused on variational algorithms and differentiable quantum computing. For the broadest job market impact in 2026, learn Qiskit first. Add PennyLane if you are targeting QML specifically.
Do I need to know Python to use Qiskit or PennyLane?
Yes. Both frameworks are Python libraries, so comfortable Python skills are a prerequisite for either. You do not need to be an expert, but you should be comfortable with functions, classes, NumPy arrays, and importing libraries before you start. If you are also targeting PennyLane for quantum machine learning, familiarity with a machine learning library such as PyTorch or JAX will help, because PennyLane integrates directly with them.
Qiskit vs Cirq: how is that different from Qiskit vs PennyLane?
Cirq is Google's quantum framework, built for fine-grained control over circuits on Google's hardware and for researchers who want low-level access to gates, moments, and device topology. The Qiskit vs Cirq decision is about which hardware ecosystem and level of control you want, both are general-purpose circuit frameworks. The Qiskit vs PennyLane decision is different: PennyLane is specialized for differentiable programming and quantum machine learning rather than being a general hardware SDK. If you want general quantum programming, the realistic choice is usually Qiskit or Cirq; if you want QML, it is PennyLane.
Is PennyLane good for beginners?
PennyLane is approachable if you already understand quantum circuit basics and come from a machine learning background, because its differentiable, functional API feels natural to anyone used to PyTorch or TensorFlow. For an absolute beginner to quantum computing, Qiskit is usually the gentler on-ramp because it has far more beginner tutorials, a larger community to answer questions, and more courses that teach the fundamentals from zero. A common path is to learn the fundamentals in Qiskit, then move to PennyLane when you start doing variational algorithms or QML.