Tool Suggestions
Giotto Alto and Solo: Switzerland’s Sovereign Reasoning Models, Free Chat and Single-GPU Running
Giotto.ai SA, a Lausanne-based AI lab, announced Giotto Solo and Giotto Alto on October 27, 2026: a new family of Swiss-developed reasoning models designed to run across public cloud, dedicated cloud and compatible on-premises hardware. The pitch is sovereignty made practical — advanced AI that people and organisations can use on their own terms, with no dependence on US infrastructure.
What launched
Giotto Alto is made available for free to everyone, through a free anonymous chat, and it runs on eight B300 GPUs. Giotto Solo is billed as the smartest model that runs on a single GPU. Alongside the models, Giotto launched apps for iOS and Android. The company’s mission, in its own words, is to make advanced AI a public good: efficient, portable and under users’ control.
What ‘sovereign’ means here
Sovereignty, for Giotto, is a deployment choice, not just a principle. There are three options: European shared cloud, dedicated cloud in Switzerland or Europe, and on-premises hardware. The target audience is enterprises, governments, institutions and individuals that want to keep contracts, internal records and other sensitive material inside an environment they control. The technical strategy is unusual: instead of building one vast model in the mould of OpenAI, Google or Anthropic, Giotto built a system of smaller specialist models that work together, which is meant to reduce the computing burden on any one task and make the product cheaper and easier to deploy. The firm says one server can be enough to run the system, which could make on-site use realistic for companies that do not want to send data to external providers.
The track record — and the gap
Giotto was founded in Lausanne in 2017 by EPFL mathematics alumni, and it has spent years chasing a benchmark rather than shipping a conventional product. In the ARC-AGI competitions, a set of reasoning puzzles that has become a prestige test in AI, its results placed 11th in 2024 and second in 2025 among roughly 1,500 teams. That is a serious reasoning pedigree.
The honest caveat comes from independent testing. Switzerland’s NZZ tested Giotto and found it could answer knowledge questions sensibly, help with drafting and talk through a PDF sewing guide, but that it was less convincing when asked to build a simple game than Google’s Gemini and still struggled with a hiking-route planning task. In other words, the promise is real, but the gap between benchmark prestige and everyday performance is still visible. Pricing for dedicated and on-premises deployment was not detailed in the release either, so the free chat is the way to judge Alto today, and the single-GPU claim is the thing to test for Solo.