Kunoichi-7B
| Property | Value |
|---|---|
| Parameter Count | 7.24B |
| Model Type | Text Generation |
| Architecture | Mistral-based Transformer |
| License | CC-BY-NC-4.0 |
| Tensor Type | BF16 |
What is Kunoichi-7B?
Kunoichi-7B is an advanced language model created through a SLERP merger between Silicon-Maid-7B and an unreleased "Ninja-7B" model. It represents a significant achievement in balancing roleplay capabilities with enhanced cognitive abilities, demonstrating impressive performance across multiple benchmarks including MT-Bench (8.14) and MMLU (64.9).
Implementation Details
The model is built on the Mistral architecture and supports context windows up to 8k tokens, with experimental support for 16k using NTK RoPE alpha of 2.6. It implements both custom and Alpaca prompt formats, making it versatile for various applications.
- SLERP merger architecture combining roleplay and cognitive capabilities
- 8k standard context window (16k experimental)
- BF16 tensor format for optimal performance
- Compatible with multiple prompt templates
Core Capabilities
- Strong performance on MT-Bench (8.14) - comparable to commercial models
- High EQ-Bench scores (44.32)
- Excellent MMLU performance (64.9)
- Enhanced logical reasoning capabilities (Logic Test: 0.58)
- Optimized for roleplay while maintaining general intelligence
Frequently Asked Questions
Q: What makes this model unique?
Kunoichi-7B stands out for its exceptional balance between roleplay capabilities and cognitive performance, achieving benchmarks that rival commercial models while maintaining strong character adherence in roleplay scenarios.
Q: What are the recommended use cases?
The model excels in general-purpose tasks, roleplay scenarios, and applications requiring both creative and analytical capabilities. It's particularly well-suited for SillyTavern implementations and can be used with various prompt templates.
