You can deploy this model and start earning money today!

lucyknada/
CrucibleLab_M3.2-24B-Loki-V1.3-exl3

Name: CrucibleLab_M3.2-24B-Loki-V1.3-exl3
Availability: InStock

CrucibleLab/M3.2-24B-Loki-V1.3

quantized

apache-2.0

Views

No views yet

exl3 quant

check revisions for quants

The Team

Steelskull

Paramancer

HuggingFace Ko-fi

Tarek

Tarek07

Mad-Lad

HuggingFace Ko-fi

Darkhn

Dataset Chronicler

HuggingFace Ko-fi

🤝 Valued Partners

Model Information

M3.2-24B-Loki-V1.3

Mistral 3.2 Finetune 24B Parameters V1.3

A special thank you to Nectar.ai for their generous support of the open-source community, which has been instrumental in helping me cover the RunPod expenses for this project.

This model has been meticulously trained on a specialized, 370 million token dataset, curated specifically for high-quality role-playing. The dataset is built upon a foundation of well-established worlds and lore, providing the model with deep knowledge across a wide array of genres.

Please use the normal mistral template

Core Themes & Genres:

The dataset's strength lies in its thematic diversity, enabling rich and immersive storytelling experiences in the following areas:

Fantasy (High Fantasy, Urban Fantasy)
Anime & Manga
Sci-Fi (Cyberpunk, Space Opera)
ERP (Erotic Role-Play)
Grimdark & Post-Apocalyptic
And many more niche settings...

Dataset Processing Pipeline:

To ensure the highest quality data, our dataset was refined through a rigorous, multi-step processing pipeline. Each step systematically cleans and enhances the data, resulting in a more coherent and reliable training base.

Processing Pipeline Overview

Step 1: Convert JSON to JSONL
Step 2: Normalize Unicode Characters
Step 3: Cleanup Separators (---, **)
Step 4: Fix OOC Misattribution
Step 5: Remove Standalone Names/Roles
Step 6: Clean Enclosing Asterisks
Step 7: Trim Last User Turn
Step 8: Fix Consecutive Assistant Turns
Step 9: Deslop Tool (Filter Content)

Recommended Sampler Settings

Static Temperature: 0.7 - 1.0

Min P: 0.02 - 0.03

DRY:

- Multiplier: 0.8

- Base: 1.75

- Length: 4-6

Good Starting Templates & Prompts

Hamon v1 → by @Steel > Big-picture storytelling guide with world-building focus, set dialogue/narration split, and general writing rules.

Shingane v1 → by @Steel > Simplified sysprompt based on Hamon.

Kesshin v1 → by @Steel > A Hamon rethink using a Character-focused sys prompt that tracks what characters know and how they learn things, with strict interaction rules.

Kamae TTRPG v1 → by @Steel > TTRPG Game Master framework emphasizing player agency, world consistency, and adaptive session management with mechanical integration.

Kamae lite v1 → by @Steel > Simplified sysprompt based on Kamae.

Quants

GGUF

Support & Community:

Join Discord

lucyknada/CrucibleLab_M3.2-24B-Loki-V1.3-exl3

exl3 quant

check revisions for quants

M3.2-24B-Loki-V1.3

The Team

Steelskull

Tarek07

Darkhn

🤝 Valued Partners

Model Information

M3.2-24B-Loki-V1.3

Core Themes & Genres:

Dataset Processing Pipeline:

Recommended Sampler Settings

Good Starting Templates & Prompts

Quants

Support & Community:

lucyknada/
CrucibleLab_M3.2-24B-Loki-V1.3-exl3