Nvidia announces open models for quantum computing: ten days of announcements without external verification

Tech & AI was also among this morning’s uncovered sections. The catch-up concerns an announcement and, together with it, the problem that announcement represents for anyone tasked with reporting on it.

Nvidia has presented Ising, described by the company as the first family of open-source artificial intelligence models dedicated to quantum computing. According to the company’s press release, the models are meant to support researchers and businesses in building quantum processors capable of running useful applications.

Where the fact ends and the claim begins

The fact is that the announcement exists and that the models are declared open. Everything else — the primacy, the usefulness, the performance — is the party’s position: this is corporate communication, and the declared performance has not been verified by third parties.

The distinction is not a quibble. In the quantum computing sector, the promise of “useful applications” is precisely what separates a prototype from a machine that actually serves a purpose, and in the material available to us that promise remains entrusted to those who produce the models.

Capital is moving around the announcement: Quantum Motion has announced a second closing of its Series C funding round, aimed at building fault-tolerant quantum computers made with standard silicon techniques. This too is corporate communication, reported by Tech Startups alongside other announcements from the sector.

Ten days of press releases

Ising is not an isolated case: it fits into a dense sequence. On September 3, 2026, OpenAI announced the release of the GPT-6 “Astra” model. On September 8, Anthropic’s commitments of roughly $80 billion in computing capacity were made public. Meanwhile, U.S. agencies issued warnings about the practice of model distillation — the technique by which one model is trained based on the responses of another. The review is by The Neuron and HIPTHER.

The common thread, stated by the reviews themselves, is that the performance data released by the labs is provided by the producers and is not independently verified.

Who should verify

In the United States, the issue has an institutional dimension. An executive order dated June 2, 2026, mandated the creation, by August 1, of a federal framework for model evaluation. That framework has not been made public in full, as reported by Fortune. At the same time, the creation of a dedicated congressional committee on artificial intelligence is under discussion.

The result is a precise situation: the only public tool designed to measure models externally to the companies exists as an act, but not in its complete form, while commercial announcements follow one another on a weekly basis.

What we don’t know

We don’t know under what license the Ising models are released, nor on which quantum processor architectures they were trained or measured: the company’s press release does not specify this in the material available to us. We don’t know the amount of Quantum Motion’s funding round. We don’t know the content of the unpublished portion of the federal evaluation framework, nor the date on which it will eventually be released.

As things stand, the available performance figures for these systems come from the companies that produce them.