[Oras]

> and a challenger to OpenAI

Big claim. Apart from Mistral OCR, I didn't find any of their models remotely useful.

They do have a chance to become sole AI provider for France as French are trying to break from US tech, not because their product is useful.

Sounds like a cry to raise more money, which is inline with their initial pitch [0]

[0] https://sifted.eu/articles/pitch-deck-mistral

[tensor]

Funny, I use them every day and find their models immensely useful. I prefer using technology from democratic countries. And no, I'm not French.

[jstummbillig]

It seems fairly plausible that they will become useful. I would not consider coding requiring vastly better AI models than the best we have today. The more interesting question is what advantages excess intelligence provides in the future more generally (warfare, system management, etc)

[Oras]

> fairly plausible

Based on what? is that a feeling or based on evidence?

[jstummbillig]

The trajectory all models are taking, while software development requirements continue to stay fairly level – at least I don't see why they wouldn't. We will still mostly need reliable, boring software.

I don't see why Mistral would stop getting better. It will, just more slowly. Eventually it will be good enough, and continue to be good enough.

[olejorgenb]

Their transcription (STT) models are good IMO

[sometimelurker]

> ... didn't find any of their models remotely useful.

I daily drive Ministral-3-3B-Reasoning, getting a nice 70 tok/s on my GPU. its actually much faster than Google/DDG searches and I use it a ton when coding to check methods and remember ways to implement things. Its better than the qwen stuff of the same size because its not filled with ccp bs.

[kkzz99]

>because its not filled with ccp bs

You do know that mistral distilled from deepseek, yeah?