Introducing Quantum 3xplorer (Q3), an AI-native engine for quantum materials modeling

Quantum 3xplorer

Meet Quantum 3xplorer (Q3, pronounced “Quantum Explorer”): an AI-native quantum simulation engine, written from scratch by AI agents under the direction and supervision of domain scientists. Q3 is open source as part of CAMSAI.org, an open, vendor-neutral community for advancing materials science with AI.

Mat3ra is part of the team behind the effort. The preprint describing Q3 has authors from Exabyte Inc. (Mat3ra.com), Lawrence Berkeley National Laboratory and its National Energy Research Scientific Computing Center (NERSC), the University of California at Berkeley, Stanford University, SLAC National Accelerator Laboratory, the University of Southern California, Johns Hopkins University, the Air Force Research Laboratory, and NVIDIA Corporation.

• Preprint: Quantum 3xplorer: AI-Native Engine for Next-Generation Quantum Materials Modeling, arXiv:2610.42108
• Code: open source under the Apache 2.0 license at github.com/camsai/q3
• Documentation: camsai.github.io/q3
• Press release: WebWire

What Q3 does

As a first and demanding validation case, Q3 implements plane-wave pseudopotential density functional theory (DFT):

• norm-conserving, ultrasoft, and projector-augmented-wave (PAW) formalisms
• spin polarization, Hubbard U and dispersion corrections
• forces, stresses, structural relaxation, and electronic-structure analysis

Q3 reproduces established reference implementations, runs faster on a single core in representative benchmarks, and scales across GPU-accelerated supercomputers. Its inputs and outputs follow structured, schema-validated data standards, so every result is AI-ready.

How it was built

Scientists define the physics, formulate the models, and set the criteria for validation. AI agents write, test, and optimize the code. Q3 was developed in weeks rather than years.

Algorithms are implemented from published equations. Validation against established codes uses only their numerical outputs. The repository's PROVENANCE.md describes how this is checked.

“Q3 enables AI-driven quantum theory. We designed it as foundational software infrastructure that turns new theoretical approaches into practical, high-performance tools for predicting the properties of materials with high fidelity, at the speed of AI.” — Timur Bazhirov, CEO of Exabyte Inc. (Mat3ra.com) and the preprint’s corresponding author

Try it

Use the links below to try the code:

• Containers: docker pull ghcr.io/camsai/q3:latest. Serial and MPI images are available for amd64 and arm64, and a CUDA image for amd64. Support for apttainer and podman containers is included.
• Building from source, tutorials, and examples: see the README and the documentation.
• Q3 runs unmodified Quantum ESPRESSO (pw.x) namelist inputs and reads UPF pseudopotentials, so existing inputs carry over.

Scope and what is next

Q3 is not intended to replace mature, feature-rich codes. They remain the reference implementations it was validated against.

Its layered architecture separates quantum-mechanical concepts from their numerical representations, so the framework can extend beyond plane waves to other basis sets and levels of theory. Current limitations are listed in KNOWN-ISSUES.md and planned work in TODOs.md. Contributions are welcome.

The preprint has not yet been peer reviewed, and its findings may change.

Cite

Biryukov, V. et al. Quantum 3xplorer: AI-Native Engine for Next-Generation Quantum Materials Modeling. arXiv:2610.42108 (2026). DOI: 10.48550/arXiv.2610.42108.

Links

As below:

• Preprint: arxiv.org/abs/2610.42108
• Code: github.com/camsai/q3
• Documentation: camsai.github.io/q3
• Data standards: github.com/camsai/standards
• Press release: webwire.com
• CAMSAI: camsai.org