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

Nearly 4000 members strong πŸ’ͺ Now with more channels! A hub for users and makers alike!


BeaverAI proudly presents...

Skyfall 36B v2

Skyfall v2 is an upscaled version of Mistral Small 2501 with continued training for creativity and RP.

image/png

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.
  • To commercial hosters of my models: If you profit off someone's work, kindly consider contributing to the cause rather than turning a blind eye to those who provide value and are in need. A subscription/donation to my KoFi would be greatly appreciated!

Supported Chat Templates

  • Mistral v7 Tekken (highly recommended)
  • Metharme (not recommended)
  • Alpaca (may be interesting, especially for cyoa / story)

Description

Creativity, good writing style, good instruct, chain of thought capability, mathematics understanding, and solid tool use performance... This model is peak! This will be my new daily model over all the 70Bs I have used.

Skyfall v2 is without a doubt my favorite model I've ever managed to run locally, bar none

skyfall is kinda nuts i am quite impressed

The biggest stand out for me is how good Skyfall handles size differences especially. It actually beats all of the 70b's I have used for descriptions of how the character worked around our size difference.

I played with the Skyfall 3bit model, taking a new character card with which I had not previously RP'd, and damn, it was so alive! The character's speech was conceptually correct, not as dry as 24b, there was a comedy tag and damn I really laughed in places. I really liked it, maybe it was the specific prompt that played great together with Skyfall.

Seriously though, Skyfall is just insanely good for some reason

How did you make skyfall so good

Links

image/gif

Downloads last month
155
Safetensors
Model size
36.9B params
Tensor type
BF16
Β·
Inference Providers NEW
This model is not currently available via any of the supported Inference Providers.
The model cannot be deployed to the HF Inference API: The model has no library tag.

Model tree for TheDrummer/Skyfall-36B-v2

Quantizations
14 models