Back home

Giotto Alto

x+web 0.88
Visit source

Intro

Giotto Alto is the flagship reasoning model from Swiss AI lab Giotto (Giotto.ai SA, Lausanne), launched on 2026-10-27 together with Giotto Solo, a model designed to run on a single GPU. The pitch is European sovereign AI.

Giotto Alto is free for everyone, hosted on European cloud infrastructure (data stays in Europe, no US cloud dependency), and ships with iOS and Android apps plus a free anonymous chat that needs no account. Website: giotto.ai; chat: cloud.giotto.ai. The official @giotto_ai launch post drew ~790 likes and 134k+ impressions and became an X trending news story.

Explain

Key points

  1. Two models:
    • Giotto Alto — flagship for demanding reasoning and enterprise workloads; runs on eight B300 GPUs.
    • Giotto Solo — billed as the smartest model that runs on one GPU, for on-prem use.
  2. Self-reported results (Alto): AIME 2026 100%, GPQA-Diamond 91.9%, Humanity's Last Exam 52%, Math-500 99.8%. Solo: GPQA-Diamond 88.3%, HLE 37.5%. Independent evaluation is pending.
  3. Three deployment options: European shared cloud (free chat; flat-fee all-you-can-use API limited to one concurrent request, with token packs for more concurrency), dedicated cloud in Switzerland/Europe (monthly), and on-premises via partner integrators (licensed per GPU).
  4. Sovereignty: GDPR-aligned, no dependency on US cloud infrastructure; aimed at enterprises, governments, institutions and individuals.
  5. Lineage: CEO Aldo Podestà confirmed it inherits some Kimi weights, with proprietary retraining on top.

Why this radar term

A newly named flagship model with an official launch, press release and X trending story. The launch post landed at the very end of the previous window and was not captured then; momentum continued in this window.

Usage

How to use it

  1. Web: open cloud.giotto.ai for free, anonymous chat with Giotto Alto
  2. Mobile: install the official Giotto app from the App Store or Google Play
  3. Developers: use the flat-fee shared-cloud API (one concurrent request); buy token packs for multi-concurrency
  4. Enterprise / government: request a dedicated Swiss/European cloud instance, or on-prem deployment through partners (Solo on a one-GPU workstation, Alto on multi-GPU servers)

Good fit

  • European organizations with data-residency, GDPR or sovereignty requirements
  • Individuals who want to try a strong reasoning model for free

Caveats

  • Benchmarks are self-reported; parameter counts are not disclosed
  • Partly built on Kimi weights — check licensing and commercial terms on the official site