8x22B?
#6
by
bdambrosio
- opened
This model is my goto. For my purposes (biomed - reasoning over research pdfs) it beats everything, even dbrx and command-r-plus. Neither can stay coherent over long-form context + long-form output as well.
So... Smaug-M8X22B? (or 141b A35, in the new HF terminology)?