Giotto launches Alto and Solo, sovereign AI reasoning models
A Swiss lab's first model family runs on European cloud or your own hardware, with Alto scoring 91.9% on GPQA-Diamond, but its licence bans publishing independent benchmarks.
Giotto, a Swiss lab founded by EPFL mathematics alumni, launched its first model family on 27 October: Giotto Alto and Giotto Solo, both built to run on European infrastructure rather than a US cloud.
Alto is the flagship model, running on eight B300 GPUs. It’s free to use through Giotto’s own chat app and cloud.giotto.ai; a shared-cloud API tier is flat-fee but capped at one request at a time, with dedicated cloud and on-premises licensing priced separately. Giotto’s own benchmark page puts Alto at 100% on AIME 2026, 91.9% on GPQA-Diamond and 52.0% on Humanity’s Last Exam. Solo is built to run on a single GPU for teams that want the model on their own hardware, and scores 96.6% on AIME 2026.
Why it matters: if data residency rules or a client contract rule out a US-hosted model, Giotto gives European teams a reasoning model they can run on their own GPU instead of trusting someone else’s cloud.
The caveat: Giotto’s end-user licence bans reverse engineering the model and bans benchmarking it for publication without the company’s written consent. The numbers above are Giotto’s own, and nobody outside the company can publish an independent check on them without asking first.