Fair enough, though I think it's rather unlikely that they'd be lying about their work on a big model, even if one might be reasonably skeptical if the model will be any good.
I think it's rather important for them that they at least keep attempting to build frontier models, because if they give up on that, then governments will no longer see them as a strateigic asset and instead just see them as a regular business.
I think that for better or worse, Mistral wants the buy-in from governments to scale up, and even if they have faltering outcomes, it's better to have that, and be able to say "look, this is what we tried, we failed due to lack of compute. We need more money for more compute."
I think it's rather important for them that they at least keep attempting to build frontier models, because if they give up on that, then governments will no longer see them as a strateigic asset and instead just see them as a regular business.
I think that for better or worse, Mistral wants the buy-in from governments to scale up, and even if they have faltering outcomes, it's better to have that, and be able to say "look, this is what we tried, we failed due to lack of compute. We need more money for more compute."