@GergelyOrosz @antirez @matteocollina
unlike most people in replies think mistral people really work hard, like I had a former colleague there working 12-15 hours a day, their researchers too
they sell models and finetuning-as-a-service to European companies like banks etc because those companies move slow due to compliance and don't need sota
Users welcome Mistral AI's medium models and domain fine-tuning for European firms because it could enable batches of small specialized models for tasks like coding.
even though I don't like European way of doing things I kinda dig Mistral's approach to not build gigantic sota generalists but rather focus on medium sized models + fine-tuning-as-a-service for domains
sovereignty is the way
unlike most people in replies think mistral people really work hard, like I had a former colleague there working 12-15 hours a day, their researchers too
they sell models and finetuning-as-a-service to European companies like banks etc because those companies move slow due to compliance and don't need sota
also for the ones who think fine-tuning is irrelevant
in TRL we see only increase in usage of different trainers, (SFT, GRPO and DPO mostly) + I know many giant companies that distill large models
companies don't want to send their data to some random US company expectedly
even though I don't like European way of doing things I kinda dig Mistral's approach to not build gigantic sota generalists but rather focus on medium sized models + fine-tuning-as-a-service for domains
sovereignty is the way
@mervenoyann I think it would be great to see "litters" of models released as batches of 10 very small specialized models. Coding obviously would be one, maybe a handful of specialized PhD's, some utility models, and an orchestrator.