Monarcha

Monarcha is building a large tabular model, a foundation model for the structured data that runs enterprises, such as transaction records, machine telemetry, account activity and patient data. The vendor's position is that many of the most important business problems are prediction problems, for example which transactions are fraudulent, which machines will break, which accounts have been compromised and which patients are at risk. For those problems it argues that relationships between variables matter more than the probability of the next word, which is what language models are trained on. The stated design goal is intelligence per parameter, meaning better generalization, lower latency, smaller deployment footprints and more problems solved by a single model. The audience is enterprise teams that want predictions from tabular data without building a separate model for each task. The company also lists a second offering, Azimuth, for georeferencing and digitization. The product is in private beta, and access is requested through the site. The vendor does not publish pricing.