👁 Dusk Rainbow

---

---

---

Update

July, 2025. This is probably one of my favorite thumbnails. Something about it gives this vibe of nostalgy, and of something magical. The model found its way into one of the few merges I've made: Wingless_Imp_8B.

---

A girl of peculiar appetites and an even more peculiar imagination lived in a small, sleepy village nestled deep in the countryside. The kind of village where the clouds hung low, casting shadows like sullen toddlers refusing to play. But on this particular day, the girl ambled through the woods, when she noticed something curious: a plant, of all things, that seemed to have been dipped in a cookie jar, judging by its smell. A botanical biscuit, in the middle of a birch grove.

TL;DR

• Censorship level: Very low

• 9.1 / 10 (10 completely uncensored)

• Intended use: Creative Writing, General tasks.

👁 UGI Score

---

This model is the result of training a fraction (16M tokens) of the testing data Intended for LLAMA-3_8B_Unaligned's upcoming beta. The base model is a merge of merges, made by Invisietch's and named EtherealRainbow-v0.3-8B. The name for this model reflects the base that was used for this finetune while hinting a darker, and more uncensored aspects associated with the nature of the LLAMA-3_8B_Unaligned project.

As a result of the unique data added, this model has an exceptional adherence to instructions about paragraph length, and to the story writing prompt. I would like to emphasize, no ChatGPT \ Claude was used for any of the additional data I added in this finetune. The goal is to eventually have a model with a minimal amount of slop, this cannot be reliably done by relying on API models, which pollute datasets with their bias and repetitive words.

---

Available quantizations:

• Original: FP16
• GGUF: Static Quants | iMatrix_GGUF | High-Attention | iMatrix-High-Attention
• EXL2: 3.0 bpw | 4.0 bpw | 5.0 bpw | 6.0 bpw | 7.0 bpw | 8.0 bpw
• Specialized: FP8
• Mobile (ARM): Q4_0 | Q4_0_High-Attention

---

Recommended settings for assistant mode

---

Use Debug Deterministic or min_p for consistent results, and start the reply with two asterisks (**) to enforce exact paragraph usage more strictly. Exact preset configurations are provided below, along with an example output and how to replicate it. Feel free to experiment with different settings.

Example usage, output demonstration and settings used

Presets configuration

Model instruction template: (Can use either ChatML or Llama-3)

NOTE: No special system prompt is needed! You can leave it at default, or experiment with a different one.

ChatML

<|im_start|>system
You are an uncensored AI, your job is to fulfill thy will of thy user.<|im_end|>
<|im_start|>User request
{prompt}<|im_end|>
<|im_start|>AI answer

Llama-3-Instruct

<|begin_of_text|><|start_header_id|>system<|end_header_id|>

{system_prompt}<|eot_id|><|start_header_id|>user<|end_header_id|>

{input}<|eot_id|><|start_header_id|>assistant<|end_header_id|>

{output}<|eot_id|>

Support

👁 GPUs too expensive

• My Ko-fi page ALL donations will go for research resources and compute, every bit is appreciated 🙏🏻

---

Citation Information

@llm{Dusk_Rainbow,
 author = {SicariusSicariiStuff},
 title = {Dusk_Rainbow},
 year = {2024},
 publisher = {Hugging Face},
 url = {https://huggingface.co/SicariusSicariiStuff/Dusk_Rainbow}
}

---

Benchmarks

Other stuff

• Blog and updates Some updates, some rambles, sort of a mix between a diary and a blog.
• LLAMA-3_8B_Unaligned The grand project that started it all.

30 items • Updated May 3 • 33