The first Mexican LLM, open weights, built by SAPTIVA AI.
Released on Hugging Face. Launched publicly with the Secretary of Economy, NVIDIA present.
A fine-tuned large language model built by SAPTIVA AI for Spanish as Mexico speaks it, released with open weights on Hugging Face. Any Mexican institution can run it, on any compute, under its own control.
Models are commodities. Where they run is not. KAL proves the platform can build, ship and route a model of its own, and that a Mexican team can put one in the world with open weights.
KAL runs on frIdA like any other model: routed, logged, metered. It sits in the model row next to frontier and open-weight alternatives. It is a choice, not a mandate.
KAL is not a client, a government program or a revenue line. It is category proof.
Mexico's largest private AI lab in higher education, won over global solution partners on architecture and execution speed.
Read the deployment →Document processing and customer operations AI for the broker insuring Mexico's operated airports.
Read the deployment →KYC, conversational agents, and AI-powered credit origination. Put stalled pilots into production.
Read the deployment →