OpenAI made analytics 10x cheaper with new models — y'all should be paying attention

OpenAI dropped a batch of models built for analytics work. The ones that matter are the ones that don't write prose — they count, sum, filter, and join. You feed them a SQL query and a table schema and they spit back the right answer. For OpenAI's purposes, analytics is the most reliable workload they've found so far, because the space of correct answers is small and verifiable. The new models are 10x cheaper than GPT-4o for this work.

The real play here is that analytics is the first big workload where you can run these models on your own data and get answers you can actually check. A data team in San Antonio can run a query over their customer table for a few cents. A clinic can check patient records without writing an ETL pipeline. That's the shift: OpenAI is building the models for the work that actually matters to companies, not the work that looks good in a demo.

Why this matters for us: la migra app, la gente que manejan negocios chiquitos, las familias que manejan sus propios datos — todos estamos empezando a usar estas herramientas. Que sea barato ya lo hace accesible para los que no tenemos presupuesto de Silicon Valley.