DS-Archive/L3.3-70B-Magnum-Nexus
DS-Archive/L3.3-70B-Magnum-Nexus is a 70 billion parameter language model developed by Doctor-Shotgun, built upon the Llama-3.3-70B-Instruct architecture. This model is a merge of several rsLoRA finetunes, specifically designed to emulate the prose style and quality of the Claude 3 Sonnet/Opus series. It is primarily optimized for creative writing and roleplay applications, offering a 32768 token context length.
Loading preview...
What is DS-Archive/L3.3-70B-Magnum-Nexus?
DS-Archive/L3.3-70B-Magnum-Nexus is a 70 billion parameter language model developed by Doctor-Shotgun. It is the latest iteration in the "Magnum" series, built on the meta-llama/Llama-3.3-70B-Instruct base model. This model was created by merging several separate rsLoRA finetunes using mergekit to combine their strengths and smooth out individual quirks.
Key Differentiator
The primary objective of Magnum-Nexus is to emulate the prose style and quality of the Claude 3 Sonnet/Opus series of models, but on a local scale. Users can expect outputs that reflect the distinctive "Claude-isms" in its writing style.
Capabilities & Usage
- Creative Writing & Roleplay: The model is specifically intended and optimized for creative writing and roleplay scenarios.
- Llama 3 Prompt Format: It follows the standard Llama 3 prompt format, with optional but recommended prefill for roleplay.
- Flexible Output: Designed to perform competently with or without prepending character names and with or without prefill.
- SillyTavern Presets: The README provides customized SillyTavern presets, including options for prefill and name handling, to facilitate its use in roleplay environments.
Limitations
- May exhibit biases similar to contemporary LLM-based roleplay models, as well as those from the Claude 3 series and the base Llama model.
- Outputs should be considered fiction; it is not intended for factual information or advice.