Hugging Face Overview
L3.3-Cu-Mai-R1-70b: Enhanced Creative Expression
Developed by Steelskull, L3.3-Cu-Mai-R1-70b is a 70 billion parameter model that represents a significant evolution in the L3.3 series, building upon the original San-Mai. It introduces a distinct "flavor" in prose and overall vibe, while maintaining strong adherence to prompts.
Key Capabilities & Features
- Model Composition: Integrates specialized components via the SCE merge method, including:
- EVA-LLaMA-3.33 for core capabilities and creative expression.
- Euryale-v2.3 and Cirrus-x1 for enhanced reasoning and coherence.
- Anubis-v1 for detailed scene descriptions.
- Negative_LLAMA for reduced bias and balanced responses.
- Prose Quality: Users praise its exceptional prose quality and natural dialogue flow.
- Prompt Adherence: Demonstrates strong adherence to user prompts.
- Reduced Repetition: Offers improved coherency and reduced repetitive outputs.
- Custom Base: Built on the entirely custom Hydroblated-R1 base model, engineered for stability and performance.
When to Use This Model
- Creative Writing: Ideal for generating engaging narratives and creative content.
- Roleplay: Excels in scenarios requiring natural dialogue and detailed scene descriptions.
- Prompt-Driven Tasks: Suitable for applications where strict adherence to prompt instructions is crucial.
Recommended Settings
- Static Temperature: 1 - 1.05
- Min P: 0.02
- DRY Settings (optional): Multiplier 0.8, Base 1.75, Length 4.
This model is part of a three-model series, offering a unique balance of creative expression and reliable performance, with reasoning capabilities activatable through proper prompting.