Anubis-Pro-105B-v1-GGUF

Join our Discord! https://discord.gg/Nbv9pQ88Xb

4000 members strong 💪

---

BeaverAI proudly presents...

Anubis Pro 105B v1 🐩

Special Thanks

• Thank you to each and everyone who donated and subscribed in Ko-Fi to make our venture a little bit easier.
• I'm also recently unemployed. I am a Software Developer with 8 years of experience in Web, API, AI, and adapting to new tech and requirements. If you're hiring, feel free to reach out to me however.

Supported Chat Template

• Llama 3 Chat for RP and Instruct
• Alpaca for Story Adventure

Description

An upscaled version of Llama 3.3 70B with 50% more layers. Finetuned further to make use of its new layers.

I'm really liking it so far. I haven't noticed any slop, the writing is good, very creative. (I know it's an overused term, but compared to other L3.3 finetunes, it really does feel so.). Definitely deserves a release. I've seen many unique generations in an hour that I've never seen before with other finetunes.

yea it writes like abliterated 3.3, follows my intended writing style nicely

I think overall this feels like a better Behemoth to me. It has a lot of elements of its emotional intelligence, ability to read between the lines and creativity, but without as much slop and with much better character adherence and prompt following. Also with fewer parameters, so it's easier to run too!

After playing around with the new Anubis upscale for a few hours I've gotta say it's my new favourite model so far. It's a bit heavy to run, but welp.

It's a great model and there's a notable intelligent jump over the base Anubis, and many other 70B Llamas I've tried. It mainly feels like an expanded ver of L3.3

Anubus Pro 105B is fantastic, I pushed it to almost 80K context and it still was pretty resonable. 0.75 temp (temp last), 0.2 smoothing factor, 2 smoothing curve, 0.01 min-p, 4 dry_mult, 1 dry_allowed_length, 3 dry_base.

I've been playing with it and it's surprisingly good! It has that... emotional intelligence that you get with 123B, but keeps that L3.3 prompt adherence that keeps characters from drifting. The best of both worlds. Even if it's a little slower, that's worth it. Running at temp 1, min_p 0.02, + Llamaception

Links

• Original: https://huggingface.co/TheDrummer/Anubis-Pro-105B-v1
• GGUF: https://huggingface.co/TheDrummer/Anubis-Pro-105B-v1-GGUF
• iMatrix (recommended): https://huggingface.co/bartowski/TheDrummer_Anubis-Pro-105B-v1-GGUF