Update Merge?

#8
by isr431 - opened

I noticed that several models used in your merge have recently received updates. Have you considered recreating the merge using these newer versions? Specifically, Lyra v4 offers substantial improvements over v1, and in my experience, Magnum v4 has shown excellent performance.

Also, I'm curious about your decision to include Mistral Nemo Bophades in the merge, as it's typically not a RP model. Could you share your reasoning behind this choice?

I noticed that several models used in your merge have recently received updates. Have you considered recreating the merge using these newer versions? Specifically, Lyra v4 offers substantial improvements over v1, and in my experience, Magnum v4 has shown excellent performance.

Also, I'm curious about your decision to include Mistral Nemo Bophades in the merge, as it's typically not a RP model. Could you share your reasoning behind this choice?

Hey, didn't know about the new Magnum, that one looks promising, but I need to check how it fares on longer contexts. As for Lyra v4, I heard mixed reviews about it, however, I should definitely try doing a new updated NemoMix-Unleashed, as per your request. It's on my “to-do” list for sure. Thanks for the reminder and recommendations!

I find Bophades to be a general “smarts” model, boosting overall intelligence as well as the roleplaying capabilities in that sense that the datasets used for it make it act more “human” if we ask people to act like one (which is precisely what I do with my system prompts). Also, all Nbeerbower's models works surprisingly well on higher contexts.

Sign up or log in to comment