Tokyo (SCCIJ) – Companies from Switzerland are pushing the boundaries of artificial intelligence, whether through efficient reasoning models or more accurate weather forecasts.
Leading in solving puzzles
The Swiss startup Giotto.ai currently leads the global ARC Prize competition that measures how close AI systems are to human-like reasoning through a series of visual puzzles. The Abstraction and Reasoning Corpus (ARC) is a benchmark for measuring progress toward Artificial General Intelligence.
At the time of writing, Giotto.ai reported that its model solved 27.08% of the puzzles. Popular large language models such as Grok 4 and GPT-5 are not very good at solving puzzles because this task requires higher levels of reasoning. Giotto also claims a markedly lower cost per task than larger AI labs. The current leg of the ARC competition ended on November 3.
Several takeover offers
The Lausanne-based startup, launched in 2017, is seeking new funding at a valuation above 1 billion dollars. Giotto has hired the Swiss investment bank Lazard to help raise more than 200 million dollars, Reuters reported. The company told prospective investors that the new capital would finance additional AI research and the development of commercial prototypes for enterprise and government clients. Giotto already received takeover offers from the USA.
The Swiss research institute Lab42 in Davos is also pursuing artificial general intelligence by supporting programmers worldwide working on their own projects to solve the ARC tests. In 2024, one of its teams set a new world record, solving 34% of the ARC tests in an unofficial challenge. “If Giotto.ai’s AI were to win the ARC Prize over the American giants, it would be a remarkable achievement for Switzerland,” the Swiss news agency Swissinfo commented.
Weather forecasting with AI
Switzerland is also involved in the race to develop the best AI weather model. In July, Swiss startup Jua introduced its model Earth Physics Transformer 2. According to Jua, it outperforms Microsoft’s Aurora and other leading conventional models in forecasting all relevant weather parameters.
Traditional weather models are based on an understanding of atmospheric physics and use physical equations. However, AI systems such as the Swiss model take a different approach: “They learn certain patterns from historical weather observations that help predict the weather for the next few days,” Sebastian Engelke, a weather model expert at the University of Geneva, told the Swiss daily paper Tagesanzeiger.
For Marvin Gabler, CEO of Jua, weather forecasting is “at a fundamental turning point.” Jua’s AI model would perform significantly better at predicting hurricane intensity than previous AI models and even classic systems. The Swiss model excels in areas where reaction speed, spatial resolution, and prediction accuracy are crucial. According to Gabler, the startup’s model is superior due to its innovative learning methods, which are better suited to weather forecasting.
Text: Martin Fritz for SCCIJ
