AiCybr

Q.ANT Open-Sources Photonic Computing SDK With CPU Simulation and Python/C APIs

By AiCybr Editorial · Published September 23, 2026

Q.ANT released an open-source software development kit for its photonic computing platform on September 23, 2026. The toolkit provides Python and C/C++ interfaces, documentation, examples and a CPU backend that lets developers build and test workloads without access to Q.ANT photonic hardware.

The official toolkit repository is available under the Apache License 2.0. Its CPU-backend mode removes the hardware-driver requirement for local development, while the hardware path targets Q.ANT Native Processing Units (NPUs). Q.ANT says access to its photonic hardware will expand through cloud partner IONOS and on-premises Native Processing Server installations in the coming months.

For developers evaluating photonic acceleration, the release creates a practical software entry point before broad hardware access. Code can be developed against the same API on a conventional CPU, then moved to Q.ANT hardware when an appropriate system is available.

What the SDK includes

The toolkit repository documents two execution paths:

Path Hardware requirement Primary use
Default backend Q.ANT hardware and corresponding drivers Execute workloads on Q.ANT NPUs
cpu-backend Standard CPU; Q.ANT drivers are not required Development, simulation and testing without photonic hardware

Python users can install release wheels or build from source. C/C++ users can install the supplied Debian package or build the shared library from the Rust-based repository. The project uses Git submodules when building from source.

Q.ANT also published a separate examples repository. The company says the examples include comparisons between conventional neural networks and networks using native photonic operations, with parameter count, accuracy, functional approximation and training loss among the measured dimensions.

Why CPU simulation matters

Specialized accelerators face a software-access problem when developers cannot readily obtain the hardware. Q.ANT's CPU backend lets teams inspect the programming model, integrate APIs and develop workloads on existing machines before committing to a photonic deployment.

That separation is particularly useful while hardware access is still expanding. Q.ANT's September announcement says cloud access through IONOS and additional on-premises Native Processing Server availability are planned for the coming months. The SDK itself is available now.

The simulation path provides a development environment for functional work. Throughput, latency and energy-efficiency decisions ultimately require measurements on the target photonic system.

How Q.ANT's photonic processor fits

Q.ANT positions its Native Processing Units as co-processors for AI and high-performance computing. Its second-generation platform performs selected mathematical operations in the optical domain using thin-film lithium niobate photonic integrated circuits.

The company's current product material lists up to 8 GOPS sustained throughput for nonlinear functions on its second-generation processor. Q.ANT also publishes substantially larger speed and energy-efficiency claims for selected workloads; those figures are vendor claims and depend on workload and comparison conditions, so they should be validated on target applications before capacity or infrastructure decisions.

Earlier in 2026, Q.ANT demonstrated diffusion-model and recurrent-neural-network workloads on its second-generation hardware. The new SDK is the software-side step that makes the programming interface directly inspectable and usable by external developers.

Installation paths

The repository documents release-artifact and source-build workflows. For Python, release wheels can be installed with pip; standalone source development uses maturin with the cpu-backend feature. For C/C++, the project can build a shared library with Cargo, with the same feature available for standalone CPU operation.

Developers targeting real Q.ANT hardware need the corresponding drivers. Developers using the CPU backend can work without those drivers or an NPU.

The repository's default branch is currently named latest. Build instructions and API details should therefore be checked against the release or branch selected for deployment.

Open-source boundary and deployment status

The core qant_native_computing_toolkit repository identifies its license as Apache-2.0, providing a permissive base for inspecting, modifying and integrating the SDK under that license's terms. Q.ANT maintains a separate public examples repository for reference workloads.

Hardware availability is at a different stage from software availability. The SDK is publicly downloadable now; Q.ANT says broader cloud and on-premises hardware access will open over the coming months. That distinction matters for teams planning evaluation schedules: software integration can begin immediately, while real-hardware benchmarking depends on access to an NPU-backed environment.

Bottom line

Q.ANT's SDK gives photonic computing a conventional developer workflow: public source code, Python and C/C++ interfaces, CPU-based development and a path to the vendor's accelerator hardware. The most useful near-term capability is the ability to evaluate the programming model without owning a photonic system. The larger performance and efficiency case will depend on reproducible measurements on real hardware as access expands.

Sources

Continue reading

AI SPARSEUP Brings Apache-2.0 Sparse Retrieval to ModernBERT

Linkup released SPARSEUP for learned sparse retrieval, with a ModernBERT backbone, Apache-2.0 licensing, 56.4 nDCG@10 on BEIR-13 and sub-millisecond Seismic retrieval results.

AI OpenClaw 2026.9.4 Guide: Skills, Plugins, Cloud Workers, GPT Image 2.5 and Security

OpenClaw 2026.9.4 adds steerable skill learning from past conversations, unified plugin and skill discovery, reusable cloud workers, GPT Image 2.5 support, terminal questions and expanded command-review controls. This guide covers upgrade requirements, architecture, security and deployment impact.

AI OpenHands 1.18 Tightens Automation Permissions and Cloud Agent Controls

OpenHands 1.18 builds on the 1.17 automation overhaul with creator-only reactivation, cloud automation editing, execution-identity visibility and stricter harness registry handling. This guide explains the release changes, Agent Canvas architecture and upgrade impact.