| Property | Value |
|---|---|
| Model Size | 123B Parameters |
| License | Other |
| Quantization | 5-bit EXL2 |
| Base Model | Largestral 2411 |
What is TheDrummer_Behemoth-123B-v2.2_exl2_5.0bpw_h6?
This is an advanced finetune of the Largestral 2411 model, representing version 2.2 of the Behemoth series. It's specifically optimized for creative tasks and implements system prompt support using a hybrid format combining Metharme with Mistral's system tokens. The model features improved creative capabilities compared to its predecessors while maintaining efficient 5-bit precision.
Implementation Details
The model implements a unique frankenformat that combines Metharme with Mistral's system tokens for optimal performance. It requires specific formatting with system prompts and maintains compatibility with various frontend implementations.
- Uses EXL2 quantization at 5.0 bits per weight
- Implements hybrid prompt format for enhanced control
- Built on Largestral 2411 architecture
- Supports both classic and creative use cases
Core Capabilities
- Enhanced creative text generation
- System prompt support for better control
- Improved performance over previous versions
- Flexible prompt formatting options
Frequently Asked Questions
Q: What makes this model unique?
This model stands out for its innovative combination of Metharme and Mistral system tokens, enabling enhanced creative capabilities while maintaining precise control through system prompts. It represents an improvement over v2.1 with enhanced creative capabilities.
Q: What are the recommended use cases?
The model excels in creative writing tasks and applications requiring system prompt support. It's particularly well-suited for scenarios where both creativity and controlled output are necessary.