This!

34b CapyB for production work.

I can’t wait for the trustworthy closed sourced benchmarks. Can’t believe I’m saying that… but it’s honestly what we need.

I actually really like that they are retaining the questions! Excellent design. By showing us a few, they gain the trust of the community in terms of the question quality.

This is the pattern moving forward and I hope that is clear to the community.

This smells like leftovers…

We’ve been having “pretraining on the test set” for weeks and I’m craving something else.

34B Nous-capybara was the only model I could use reliably for complicated nlp and json output. My go to for any real work. The first, really.

I use nous-capybara for nlp processing with json output for work.