redrix/patricide-12B-Unslop-Mell-v2
redrix/patricide-12B-Unslop-Mell-v2 is a 12 billion parameter merged language model created by redrix using the NuSLERP method, combining inflatebot/MN-12B-Mag-Mell-R1 and TheDrummer/UnslopNemo-12B-v4. This model is designed with an emphasis on "anti-GPTism" influence, aiming for less conventional or more creative responses. It is primarily intended for conversational applications using the ChatML template, with early testing showing stability and viable answers.
Loading preview...
Overview
redrix/patricide-12B-Unslop-Mell-v2 is a 12 billion parameter language model developed by redrix through a NuSLERP merge of two base models: inflatebot/MN-12B-Mag-Mell-R1 and TheDrummer/UnslopNemo-12B-v4. The creator specifically chose UnslopNemo-12B-v4 for its reported "anti-GPTism" influence, aiming to produce responses that deviate from typical LLM outputs, potentially offering more unique or less predictable interactions.
Key Characteristics
- Architecture: Merged model using the NuSLERP method.
- Parameter Count: 12 billion parameters.
- Context Length: Recommended not to exceed 20,000 tokens for optimal coherency, despite a larger potential context.
- Chat Template: Primarily configured for
ChatML, thoughMetharme/Pygmaliontemplates might also be compatible. - Development Stage: Currently in early testing, with initial results indicating stability and viable answer generation.
- Quantization: GGUF quantizations are available from community contributors like MaziyarPanahi and mradermacher.
Use Cases
This model is suitable for conversational AI applications where a less conventional or more "anti-GPT" response style is desired. Developers looking to experiment with models that prioritize unique output over strict adherence to common LLM patterns may find this model useful. It is recommended for early-stage testing and exploration of its distinct response characteristics.