Changelog
The format is based on Keep a Changelog, and uses Semantic Versioning.
[1.3.0] - 2026-06-24
Added
Classical Chemistry Calculators
We added several new classical chemistry reference calculators including:
Second-order Møller-Plesset (MP2) ground-state problem calculator and reaction path problems calculator.
Coupled cluster (CCSD and CCSD(T)) ground-state problem calculator and reaction path problems calculator.
TRIM-CI (Truncated Restricted-active-space Interaction with Many-body perturbation) configuration interaction calculator.
We also added the coupled cluster singles and doubles (CCSD) as a new molecular orbital calculator.
Tensor Network Simulation
We added a new tensor network simulation module with support for both CPU and GPU backends.
GPU Acceleration
We have added GPU-acceleration to the orbital optimization Hessian (supported on Linux and MLX support on macOS) We have also added the CuQuantum state-vector estimator for GPU-accelerated exact simulation.
Optional Extras for Hardware Backends
Hardware backend dependencies (IBM, IQM, OQC, CUDA) have been restructured as pip optional extras. This means a bare
pip install qrunchno longer pulls in heavy, backend-specific dependencies you may not need. Install only the extras required for your target hardware, e.g.pip install qrunch[iqm]to get the IQM backend working. Multiple extras can be combined:pip install qrunch[iqm, oqc]. To restore the previous behaviour where all backend dependencies are included, usepip install qrunch[all].Shadow Grouper
We added a Shadow grouper that creates groups by sampling the most important Pauli strings for the observable first.
Better shot utilisation
We changed the definition of shots and added shot distributors with the new
.with_total_estimator_shots(shots=1_000_000)method. The new method sets the total number of shots and distributes them across measurement groups, ensuring better use of the shot budget on hardware. The existing.with_estimator_shots(shots=1000)continues to work as before, setting the number of shots per call to the quantum computer or simulator. But deprecated in the next release.Performance Improvements
Several general performance improvements.
Performance improvements for OO-FAST-VQE and OO-BEAST-PT2 VQE workflows.
[1.2.0] - 2026-05-22
Added
Rigetti Cepheus 108Q support via Amazon Braket.
Auxiliary energy contributions (DFT mean-field energies) for projective embedding problems.
Active electrons correlation energy property added on
GroundStateProblemCalculatorResultVerification that hardware backends function correctly on creation of a backend sampler or estimator.
Data persister support for the Configuration Interaction calculator.
Improved default for RI/Auxiliary basis sets.
VQE warm start utilities for starting VQE calculations from a previous result, a custom circuit, or a parameterized circuit.
Changed
Upgraded Qrunch to Python 3.12.
Fixed
Fixed measurement result bit ordering for IQM backends.
Fixed memory leak in RI integral evaluation.
Fixed support for Amazon Braket Hybrid Jobs.
[1.1.0] - 2026-03-03
Added
BEAST-PT2 (Second-Order Correction for BEAST-VQE)
BEAST-VQE can now be augmented with a second-order perturbative correction to improve accuracy beyond the paired-electron approximation. The correction supports pair-breaking effects, single excitations, or both, and can be enabled directly in the calculator configuration.
Noisy Memory Restricted Simulator
A new memory-restricted simulator can emulate physical quantum hardware, including user configured device data. It combines the standard memory restricted simulator with Monte Carlo sampling, to emulate hardware-like behavior based on calibration data.
GPU Acceleration
We introduced a new GPU acceleration module with initial support for accelerated orbital rotations, providing improved performance for selected workflows.
Hardware Backend Improvements
Support for IBM NightHawk architecture devices (e.g., Miami).
Optional instance selection for IBM Quantum Platform backends.
Improved robustness of IQM backends when handling missing qubits or calibration data.
Performance Improvements
Faster mapping from second-quantized Hamiltonians to Pauli operators.
Performance improvements for FAST-VQE and BEAST-VQE workflows.
Reduced logging overhead in the memory-restricted simulator.
Deprecated
Tensor Hypercontraction integrals have been deprecated as it never worked to our satisfaction and it will be removed in a future release.
Fixed
Fixed a performance issue in even-handed projective-embedding calculations.