Maybe that's the honest European answer: stop chasing the frontier and win at being the best at making models useful. But then someone else owns the factory, and Europe rents from it.

What about Ami? HQ in Paris I believe. Arguably on a frontier of their own re: JEPA / world models

Mistral is using chinese open weights internally

99% of top AI researchers that live in EU work for US frontier labs. EU AI companies cannot attract top talent. Forget about EU.

prof there must come a time when models start experiencing the law of diminishing returns in your opinion have we reached that point sometimes using the latest models feels like a regression compared to their older compatriots

Honestly the frontier gap barely registers for anyone making actual work. Mistral could close it tomorrow and I still couldn't name a single thing I'd open it for.

Esta distinción es clave para finanzas: no siempre conviene el modelo más grande, sino el que ofrece costo, latencia y desempeño verificable según la materialidad. En conciliaciones y auditoría, benchmark propio y evidencia trazable antes de integrarlo al ERP.

Mistral hasn’t pivoted away from frontier models. Arthur Mensch has explicitly said they’re working on a new open-weight model family aimed at closing the gap with the frontier.

Mistral is pivoting towards being am inference provider rather then being a frontier lab. Sadly EU misses the ambition, everyone just talks.

In my own benchmarks, Mistral was the best at translating to European languages.

Facts are meaningless. Ethan. People believe what they want to believe and period.

Yes, they have abandoned the frontier model race in favor of specialized models. Europe cannot reasonably take part in this competition; it has neither the resources nor the will to do so...

A new big model will be released in the next weeks