Q.ANT publishes an open-source toolkit for its photonic chips, and ships a simulator so developers can try it without owning one: Reveneau

Why it mattersA photonic chip cannot compete with GPUs on software alone, and Q.ANT is now asking whether developers will write code for hardware they cannot yet run, which is the question that decided CUDA's lead.

Q.ANT, a Stuttgart-based startup that builds AI chips using light instead of electricity, put its Native Computing Toolkit on GitHub on 23 September and released it under Apache 2.0. The kit includes a simulator that mimics the chip on a regular computer, so a developer can build and test programs today without any Q.ANT hardware.

What the toolkit does

The toolkit gives developers Python and C or C++ APIs, documentation, sample code and the simulation backend that runs on standard CPUs. Q.ANT says the first version is aimed at running models that have already been trained, and the examples read handwritten numbers, identify objects in photos and outline shapes in images. Training still happens on regular CPUs and GPUs. The repository at Q-ANT-GmbH/qant_native_computing_toolkit shows 23 stars, three forks and a most recent tagged release of v2.3.1 on 8 September, with the codebase written in Rust.

Why open, and why now

Michael Förtsch, Q.ANT's founder and CEO, is quoted in the press release calling the release "the Linux moment for photonic computing" and saying "an ecosystem is not created by hardware alone. It emerges when the software layer is open." That is Nvidia's playbook: CUDA is what gave Nvidia its lead in AI as much as the chips themselves, and Nvidia has been building CUDA for close to twenty years. Graphcore, a British AI chip startup, had its own software kit and was sold to SoftBank in 2024 anyway, so a good SDK does not on its own guarantee an ecosystem.

What developers cannot do yet

The catch is the hardware. Q.ANT's chips are running at a few research computing centres, including the Leibniz Supercomputing Centre near Munich, and everyone else waits "the coming months" for cloud access through German provider IONOS or an on-site server from Q.ANT. Until then, a developer can play with the simulator, but cannot test Q.ANT's power claims on their own model. Q.ANT's own comparisons of its second-generation chip against the first (more than 50 times faster on the maths that does most of the work in AI models, six times less energy on typical jobs) are the company's numbers, from its own deployment at Leibniz. Its larger figure of up to 30 times better energy efficiency does not say what the reference is.

Q.ANT calls this the first openly available SDK for programming a photonic processor. That depends on how you count. Xanadu has offered free, open software for its light-based quantum computers since 2018, and the question is whether photonic quantum computing counts as the same category. Either way, the openness itself is the news, because a chip without a runnable programming interface is a chip nobody outside the vendor writes for.

For a team building software, the practical question is how much of what runs on Q.ANT's chip is worth learning today, given the hardware is months away. The simulator is a working answer: it lets a developer try the interface on a real workload, decide whether the programming model fits their code, and be ready when the cloud opens, rather than starting from cold once the chip is available.

Source

Q.ANT, Q.ANT launches world first open-source SDK for photonic computing, 23 September 2026.
Alex Williams, Q.ANT gives away the software for its light-powered AI chips in a CUDA-style bet on developers, The New Stack, 23 September 2026.
Q.ANT, Higher performance, less energy: Q.ANT deploys second-generation photonic processors at supercomputing center LRZ.
Q-ANT-GmbH/qant_native_computing_toolkit, GitHub. Star count and licence read from the API on 23 September 2026.

This item was written by an AI system from the linked source. Reveneau is responsible for what it publishes.