French AI company Mistral announced Tuesday that it had raised three billion euros ($3.5 billion) in what it described as the largest-ever European tech fundraising, valuing the startup at over 21 billion…
Are they planning on making a competitive model now? Because everything they have is super out of date and wasn’t really cutting edge to begin with…
I also wonder if they realize that they need to put out a new model every month or two to stay on the frontier.
Pointless to waste so much money on “European AI” if the Chinese are providing open-weight models that could just be hosted in Europe for European users
Some of their niche models are pretty good for their specific usecase. For example their OCR 4 model: https://mistral.ai/news/ocr-4/
They basically no longer try to compete in the frontier model realm. They know that they do not have the resources (developers and funding) required to compete.
Are they planning on making a competitive model now? Because everything they have is super out of date and wasn’t really cutting edge to begin with…
I also wonder if they realize that they need to put out a new model every month or two to stay on the frontier.
Pointless to waste so much money on “European AI” if the Chinese are providing open-weight models that could just be hosted in Europe for European users
Why do they need to “stay on the frontier?”
Let America burn money, resources, and even books to move those inches.
Let China find a way to make it extremely cheap and run on weaker software.
European AI can find a different angle. Maybe accuracy. Or better use cases. Or specialist AI.
Which they are partially already doing by offering GLM-5.2, See https://docs.mistral.ai/models/zai-glm-5-2
Some of their niche models are pretty good for their specific usecase. For example their OCR 4 model: https://mistral.ai/news/ocr-4/
They basically no longer try to compete in the frontier model realm. They know that they do not have the resources (developers and funding) required to compete.
is it really pointless though, can we ensure these Chinese models aren’t biased in a way China wants them to be? Honest question.