Fin fine-tunes NVIDIA Nemotron 3 Super and Nano for Fin Apex models

We’re pleased to announce the launch of our 2.0 series of Fin Apex models! We think these are the best models for customer service. They are incredibly fast and efficient to run, but even beyond that, we’re seeing an increasing, not shrinking, performance gap between our specialised models, and the general frontier. We recently did a backtest against Astra, an amazing general model, and you’d expect overkill for customer service - but it still underperforms Apex on our particular tasks. So, at least in our experience, there’s still a lot of room for domain specific customisation. We try to give some flavour of the sort of tradeoffs necessary to achieve this performance in the launch post: https://fin.ai/blog/announcing-apex-2