Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.haiqu.ai/llms.txt

Use this file to discover all available pages before exploring further.

​
May 30. 2026
v1.0.5
​
May 18, 2026
v1.0.4

​
Haiqu v1.0.4

We would like to thank our early access users for their continued feedback and feature requests, which have helped shape many of the improvements introduced in this release.

​
Load matrix product states

Haiqu now supports direct loading of matrix product states (MPS) onto quantum processors through the new mps_loading function.
mps_loading
This expands Haiqu’s portfolio of advanced data loading techniques, which already includes vector loading, block vector loading, distribution loading, and entangled manifold embedding (previously known as isometry encoding). To learn more, visit: https://docs.haiqu.ai/reference/data/mps_loading

​
Transpile circuits efficiently

A new multi-pass transpilation workflow automatically selects the most effective optimization strategy for a given circuit. These improvements deliver greater robustness and performance, reducing two-qubit gate depth by an additional ~37%.To learn more, visit: https://docs.haiqu.ai/reference/run/transpile

​
Visualize compressed circuit structure

The Haiqu dashboard now features dedicated circuit pages that provide users with a clearer view of circuit structures and key execution metrics, including: active qubit count, two-qubit gate depth, and estimated quantum cloud cost. Users can also inspect noise contributions across qubits and circuit layers alongside detailed circuit diagrams for deeper analysis and debugging.
compression_visualization

​
Enjoy additional improvements

Improved resource utilization for dynamical decoupling, enabling 100x faster transpilation passes in some cases. Improved observable-based readout error mitigation.
​
May 11, 2026
v1.0.3

​
Haiqu v1.0.3

​
Apply robust error mitigation

In this release we enhanced reliability of the error mitigation routines for more consistent execution and improved stability when combined with other techniques (e.g., block vector loading).
Enhanced Reliability

​
Gain quick insights with execution frontier plots

Updated quantum volume visualization to include execution frontier, providing deeper insight into circuits and their expected performance.
Quantum Volume Visualization

​
Enjoy additional improvements

Added support to log dataframe tables and render them on the dashboard, enriching experiment context and analysis.Delivered dashboard performance optimizations, resulting in faster query execution and rendering speeds across the UI.Simplified function interfaces by making IBM Credentials optional when operating with fake devices or in dry_run mode.
​
April 2026
v1.0.0

​
Haiqu v1.0.0

​
A Platform for Enterprise Quantum Workloads

Today we’re launching Haiqu v1.0, a platform built to help businesses develop and run practical quantum workloads on today’s quantum hardware.Haiqu provides a unified software layer for building, optimizing, analyzing, and executing quantum applications across leading quantum providers, including IBM Quantum, Amazon Braket, and IonQ. The platform is designed to reduce the complexity of working with noisy quantum systems while helping teams evaluate real-world quantum use cases more efficiently.

​
Make Every Quantum Dollar Go Further

Quantum computing remains expensive, especially at scale. Haiqu reduces quantum cloud costs in three ways: by rewriting circuits into shallower and more efficient forms, by improving hardware utilization through orchestration, and by increasing solution fidelity so fewer shots are required overall. Some teams have already reported 1,000× lower execution costs using Haiqu.
compute_execution_summary

​
Hardware-Aware Optimization for Real Quantum Systems

A major focus of Haiqu v1.0 is performance optimization. The platform seamlessly applies hardware-aware circuit compression, transpilation, and error mitigation techniques to improve execution fidelity on current-generation quantum processors. These capabilities help reduce circuit depth, lower gate counts, and improve solution quality without requiring teams to manually tune circuits for each backend.
quantum_systems

​
Scalable Quantum Data Loading

Haiqu v1.0 also introduces four new scalable quantum data loading and feature embedding capabilities for data-intensive applications such as quantum machine learning and optimization. By dramatically reducing the two-qubit gate depth required for data loading (by more than 350× in some cases), Haiqu enables teams to fit larger and more practical workloads onto real hardware, leaving more room for the algorithm itself to run.
circuits_table

​
Integrated Analytics and Experiment Management

To support enterprise experimentation and benchmarking, Haiqu includes integrated analytics and experiment management tools that allow teams to track executions, compare circuit performance, analyze hardware feasibility, and monitor optimization results across providers and devices.
experiment_management

​
Developer-First Workflow

Haiqu is built around a developer-first workflow with native Qiskit interoperability, Python APIs, and cloud-hosted JupyterLab environments that make it easy to start building and running workloads quickly.The platform also integrates with modern AI-assisted development workflows through a model context provider that understands commands such as “Compress this circuit” or “Apply error mitigation,” helping teams move faster from ideation to working prototypes.
developer_workflow

​
Enabling the Next Phase of Quantum Adoption

Quantum computing is entering a new phase where organizations are moving beyond theoretical exploration toward practical evaluation of business applications. Haiqu v1.0 is designed to help enterprises navigate that transition by providing the infrastructure needed to optimize workloads, benchmark hardware, and accelerate experimentation on real quantum systems.

​
Deployment

  • Flexible Deployment Options: Added support for enterprise deployment within private VPC environments, enabling secure and customizable infrastructure setups.

​
Performance Benchmarking

  • Unified Results Interface: Introduced a standardized interface for all quantum run results with built-in performance estimates.
  • Middleware Optimization Controls: Users can now enable or disable middleware optimizations to evaluate their impact on performance.
  • Pareto Analysis GUI: Added a graphical interface for visualizing trade-offs between accuracy, cost, runtime, and hardware parameters.
  • Automated Run Tracking: All runs are now automatically documented, providing a clear and auditable decision trail.

​
SDK & Metrics

  • Enhanced Proxy Performance Metrics
    • Added error bars to expectation values
    • Improved consistency of gate-level metrics
    • Expanded circuit metadata (author, job linkage, transpilation differences)
  • Improved Circuit Insights enabling diagnosis of:
    • QPU result reliability
    • Root causes of poor performance
    • Suitability of problems for hardware execution

​
QML & Simulation

  • Iterative QML Workflows: Added support for iterative training, pretraining, and deployment of QML mitigation pipelines via SDK.
  • Statevector Simulator: Introduced a high-performance simulator for debugging, development, and pretraining.
  • Device-Aware Circuit Compression: Added compression optimized for IBM heavy-hex and square architectures.
  • Advanced Data Loading: Introduced isometry-based data loading with significantly improved fidelity and UX.

​
Dashboard

  • Migrated the dashboard to React with a redesigned UI.
  • Device Management Improvements: Added enhanced device visibility and detailed device information widgets.
  • Experiment Reporting Enhancements
    • New performance widgets
    • Circuit-to-device mapping
    • Improved notebook integration
    • MCP server with access to experiment data for summarization and interpretation
  • Differentiator Widget: Highlights experiments using key Haiqu features such as mitigation, compression, and large-scale data loading.

​
Identity & Access Management

Automatically sync API keys between Dashboard and Lab environments after updates (reduces setup friction, prevents configuration issues, and makes account management faster and more seamless).

​
Billing & Credits

  • Added automatic monthly credit refresh.

​
Billing & Credits

  • Users are no longer charged for time spent waiting for the QPU vendor (while the QPU job is queued or running).

​
Performance & Execution

  • Improved decision-making workflows with clearer performance insights and benchmarking tools.
  • Enhanced transparency of circuit execution and optimization processes through richer metadata and metrics.
  • Optimized execution efficiency:
    • Reduced runtime and cost through better compilation and orchestration.
    • Improved accuracy via QML workflows and circuit optimizations.
  • Gate Metrics Inconsistencies: Fixed inconsistencies in reported gate-level performance metrics.
  • Circuit Data Representation: Corrected issues in circuit tables to ensure accurate and consistent metadata display.
​
January 2026
v0.3.0
This release introduces updates across SDK, Experiment Tracking, Dashboard, Credits & Pricing, and User Guides documentation (Jupyter Notebooks) with improvements to usability, reliability, and observability.

​
1. Pricing and credits management

We set up two pricing models:
pricing_models
Early access includes 1 hour of execution time to explore our SDK and Dashboard features. A credits usage progress bar is displayed on the Dashboard. Users receive updates about future releases and deployments through a popup message on the main Dashboard screen.
credits_tracking

​
2. Experiment Tracking: Reports

Experiment reports support LaTeX-style Markdown for scientific formatting and allow users to add artifacts via drag-and-drop, making it easy to compose structured, shareable experiment summaries.Users can log Matplotlib plots as experiment artifacts, which appear in the Dashboard as part of the experiment history. Execution result tables can also be logged as artifacts. Supported tables include: Circuit core metrics, Circuit advanced metrics, Circuit all metrics, and Circuit comparison tables.
experiment report

​
3. Security and Privacy

Haiqu is ISO/IEC 27001 compliant and uses IAM-based controls to ensure customer data is handled with strong, auditable security practices. Access is granted only on a need-to-know basis, and clear records of who accessed what and when, supporting both compliance requirements and operational transparency, are maintained.To further reduce insider risk, production access is exception-based and support access is temporary and approved, with all activity logged and monitored. When needed, we can also provide access logs to help customers investigate incidents and meet internal governance standards.
Security and Privacy
Image: Haiqu’s security and privacy practices as well as ISO certifications are documented our trust center webpage.

​
4. Method haiqu.postprocess()

  • What does it do? Applies classical heuristics (e.g., bit-flip search) to measured bitstrings to find a better solution to the QUBO problem.
  • How do I use it? Pass counts or probability distribution obtained by running circuits on any backend, and pass the corresponding QUBO problem object, then call postprocess().
post_processing method results
Table: Experiments result to compare a raw quantum results and post-processed results on a 120-qubit optimization problem

​
Dashboard—Experiment Tracking

  • Plot logging Users can now log Matplotlib (MPL) plots as experiment artifacts using the haiqu.log() method. Logged plots appear in the Dashboard as part of the experiment history.
  • Logging of tables as artifacts Users can now include notebook-generated tables and visual outputs as artifacts in experiment reports. Added logging support for:
    • Circuit core metrics
    • Circuit advanced metrics
    • Circuit all metrics
    • Circuit comparison tables
  • Reports in experiment tracking
    • LaTeX-style Markdown support (UX)
    • Ability to insert artifacts into reports by drag and drop
  • Job deletion support

​
Dashboard—Pricing Model

  • Pricing and access management
    • Defined two pricing packages
    • Added a Credits Request form and Pricing panel in the Dashboard
    • Enabled contacting support via Slack
    • Designed credits expiration flow in the SDK
    • Added Admin Panel UI for manual credit allocation
  • Credits management
    • Account creation now includes allocated credits (1h of execution time)
    • Added a credits usage progress bar on the Dashboard
    • Implemented credits consumption mechanics
    • Added SDK error messaging when credits are exhausted
    • Allows already-started jobs to finish even if credits expire during execution

​
SDK—Execution

Renamed the run method argument (backend_name → device_id) for clarity and consistency when selecting execution devices.

​
SDK—Postprocessing

Updated the haiqu.postprocess and haiqu.solve_qubo methods:
  • Improved QUBO problem definition and solution using a non-iterative method
  • Enhanced results postprocessing pipeline
  • Added detailed documentation explaining the updated workflow

​
Jupyter Notebooks

Released an updated set of Jupyter notebooks in our Jupyter Lab env to help users get started with the Haiqu SDK, reflecting the latest APIs, workflows, and examples.
  • IonQ transpilation fix: Fixed a transpilation error for IonQ devices by updating the equivalence library in the transpilation worker to support IonQ-specific gates.
  • Aer simulator transpilation fix: Fixed a transpilation failure caused by a missing operation_names attribute when using aer_simulator.