At Trifolia.cl we have no blind faith in any artificial intelligence model. We don’t fall in love with brands.
What we do is simpler and harder at the same time: test, measure, choose.
We have an internal process of continuous model evaluation. Response quality, speed, contextual accuracy, cost. Everything counts. And when we find a better one, we switch without hesitation.
When Anthropic released Claude Sonnet 4.6, the next day we were already running it in our agent. Not out of fashion or hype. We tested it because testing everything is our job. And it turned out to be good. Very good. Sonnet 4.6 won on what matters most to us: the quality of the answers our users receive. Today it’s one of the core models that power Trifolia.
This is one of the real advantages of building on AI in 2025: the pace of improvement is brutal. And if you’re paying attention, you can pass that improvement straight on to your clients.
In legal research there’s no room for a pretty mistake. A lawyer can’t show up in court with a citation that doesn’t exist or a law from another country. That’s why we choose models the way you choose a scalpel: by precision, not by brand.
Technology moves fast. We move faster.
Originally published on LinkedIn on March 4, 2026.