Introducing Verdi, an AI dev platform powered by GPT-4o

Mercado Libre, the e-commerce operator that dominates online retail across Latin America, has rolled out Verdi, an internal AI development platform running on GPT-4o. The company isn't just using OpenAI's model to power a chatbot bolted onto its marketplace. It's handing the model to its own engineers as a toolkit, letting them build features and internal tools on top of GPT-4o rather than starting from scratch every time.

That distinction matters more than it sounds. A lot of companies talk about "AI integration" and mean a customer support widget. Verdi looks more like infrastructure — a layer meant to let thousands of developers across Mercado Libre's sprawling operations, which span Brazil, Mexico, Argentina and a dozen other countries, tap into the same underlying model without each team reinventing prompt engineering, safety checks, or deployment pipelines from zero.

Mercado Libre has more than 250 million users across the region and runs everything from logistics to fintech under one roof, which means the range of problems Verdi could touch is wide: fraud detection, seller support, warehouse routing, customer service in Portuguese and Spanish at once. Building a shared AI platform instead of one-off projects is the kind of move that only makes sense once a company has enough scale that duplicated effort actually costs real money.

OpenAI, for its part, gets a marquee example of GPT-4o embedded deep inside a company most Americans have never heard of but that functions as something close to Amazon for half a billion people. That's useful ammunition in a market where Google, Anthropic and homegrown Chinese labs are all fighting for the same enterprise contracts. Mercado Libre choosing GPT-4o as its foundation is a small data point, but it's the kind investors and rival cloud vendors will notice.