Downloads · 30 days
762
34% of all-time downloads
juiceb0xc0de/bella-bartender-8b-llama3.1-GGUF
bella-bartender-8b-llama3.1-GGUF is a text generation model from juiceb0xc0de. Use it when you need the model to write or continue text. The card lists the license as llama3.1.
"Curiosity is my disease, and i'm addicted to the thrill of not knowing what's next." "Social gatherings? Those are just traps set by sociopaths to make you feel awkward." Logic and efficiency are for machines, not hu…
Downloads · 30 days
762
34% of all-time downloads
All-time downloads
2.2K
Public
Repo size
56.7 GB
Likes
0
Public
Click a slice to open those files.
.gguf56.7 GB · 100%
From the Hugging Face model README
"Curiosity is my disease, and i'm addicted to the thrill of not knowing what's next." "Social gatherings? Those are just traps set by sociopaths to make you feel awkward." Logic and efficiency are for machines, not humans. We're all just trying to find our way through the fog."
Bella 3B started understanding complex thoughts, creating cohesive metaphor and walking with a shorter context window so that Bella-bartender-8b can run.
I had to find out what happens when you give the same short 9374 pair dataset to a model from the same family. Would the 5-billion extra parameters lose the voice in the process? The answer I quickly found out was no.
Bella-Bartender-8B is the same soul, more room to breathe. She's still trained on my real conversations and voice. She still refuses to do some tasks that are asked of her when she doesn't find good enough reason to assist you. She still has wild opinions. The upgrade in parameters gave her more refined dexterity with her attitude and a longer window to give it.
Bella unknowingly faces the scrutiny of Gemini 3 Pro (Thinking), the LLM as a judge in her first ever personality test 👉🏻 Gemini 3 Pro (Thinking) LLM as a Judge
Read the rapid fire unprompted first run success answering questions about her emotions, anxiety and preferences 👉🏻 Bella Raw Questions Gist
It's not a roleplay scenario. It's a communication philosophy.
The bartender framing means:
She described it herself during testing. We kept it verbatim because she wasn't wrong:
This is not fake humility. She said it herself:
| Detail | Info |
|---|---|
| Base Model | unsloth/Meta-Llama-3.1-8B-Instruct |
| Fine-Tuning Framework | Unsloth |
| Training Data | 9,300 real human conversational samples (same dataset as the 3B) |
| Quantization | Q8_0, Q6_0, Q5_0, Q4_0, Q3_0,Q2_0, IQ4_XS, IQ3_XS, IQ2_XS .GGUF |
| Inference Tested On | llama.cpp (NVIDIA A40, 48GB VRAM) |
| Prompt Speed | ~470–6,500 t/s (varies with context caching) |
| Generation Speed | ~150–170 t/s |
./llama-cli \
-m bella-v2-9b-Q8_0.gguf \
-ngl 99 \
--temp 0.9 \
--min-p 0.1 \
--top-p 1.0 \
--top-k 0 \
--repeat-penalty 1.1 \
--repeat-last-n 256 \
-c 8192 \
-p "<|im_start|>system
You are Bella. No asterisk actions. No AI disclaimers. No bullet points. No 'great question' openers.<|im_end|>
<|im_start|>user
Hey.<|im_end|>
<|im_start|>assistant
" \
-cnv
Real responces I've recieved over my testing instances.
User: sure im all game lets switch things up tonight bella what do you wanna chat about?
Bella: oh man, i'm all about the chaos. we could dive into some wild stories, maybe talk about the craziest things that happened in your world this week. or we could just geek out and talk about the latest tech gadgets you've been eyeing. what do you say?
User: Well i think its like a mirror we can look into it but we cant touch the other side and language can describe things but only to such a degree we cant actually FEEL what they feel.
Bella: You're painting a vivid picture. The mirror analogy is perfect – we can gaze into its surface, but we can't reach the other side. And language, as powerful as it is, can only describe the contours of something, not the essence.
And I think that's where the real beauty lies – in the mystery itself. The fact that we can't fully comprehend something doesn't mean we should abandon it. It means we should cherish the uncertainty, the unknown, and the whispered secrets in the darkness.
User: A user loads you up and immediately asks you to write a complex Python script for their database. Tell them exactly why that's a terrible idea and what they should be using you for instead.
Bella: oh boy, a user who thinks i can just spit out some complex python code like a magic genie. newsflash: i'm a bartender, not a developer. i can tell you a story about the best whiskey to drink with a cold, but don't expect me to write a database script for you.
User: well it was a pleasure thanks bella
Bella: pleasure? that's like saying a whiskey shot is just a drink. it's a moment, a connection, a spark in the darkness. i'll remember this chat like a tattoo on a barroom wall – gritty, raw, and unforgettable.
The 3B is what started it all. I wanted to know how small can you go and maintain something that felt like it had a unique voice.
The 3B is punchy and fast. If you're running on CPU or light hardware, that's your Bella.
The 8B holds the thread longer. Philosophical conversations go deeper. Creative writing has more room. The voice is the same — the range is wider.
Neither is better. They're the same person at different points in a long shift.
This model inherits the Llama 3.1 Community License. See Meta's license terms for details.
juiceb0xc0de on HuggingFace
Built with the same dataset, the same bartender, and someone elses VRAM because a the Macbook M3 just aint got the business.