nmalinowski/pauper-llama3-8b

TEXT GENERATIONConcurrency Cost:1Model Size:8BQuant:FP8Ctx Length:8kPublished:Jan 1, 2026License:llama3Architecture:Transformer0.0K Cold

The nmalinowski/pauper-llama3-8b is an 8 billion parameter language model, fine-tuned from Meta's Llama 3 8B Instruct architecture. Specialized using LoRA, this model excels in understanding and generating content related to Magic: The Gathering's Pauper format. It is designed for detailed inquiries about Pauper cards, deck building, and meta-game analysis, offering a focused resource for enthusiasts.

Loading preview...

Pauper Llama 3 8B: Specialized MTG Pauper AI

This model, developed by nmalinowski, is a fine-tuned version of Meta-Llama-3-8B-Instruct specifically designed for Magic: The Gathering's Pauper format. Utilizing LoRA (Low-Rank Adaptation) fine-tuning, it has been optimized to provide in-depth knowledge and generate relevant content about Pauper.

Key Capabilities

  • Pauper Format Expertise: Deep understanding of cards, rules, and strategies within the Magic: The Gathering Pauper format.
  • Deck Building Assistance: Can help construct Pauper decks and explain card synergies.
  • Meta-Game Analysis: Capable of discussing current Pauper meta-trends and top-tier decks.
  • Content Generation: Generates responses to queries about Pauper removal spells, deck archetypes (e.g., Affinity vs. Elves), and card interactions.

Available Formats & Usage

The model is provided in both HuggingFace Transformers (full precision) for further fine-tuning and maximum quality, and GGUF quantizations for efficient local inference on consumer hardware (compatible with LM Studio, Ollama, and llama.cpp). The recommended q4km.gguf quantization offers approximately 95% of full precision quality with a 70% smaller file size, making it ideal for most users.

Limitations

  • Domain Specificity: Highly specialized for the Pauper format; performance on other MTG formats or general topics may be limited.
  • Potential Hallucinations: May occasionally generate inaccurate card names or abilities.
  • Knowledge Cutoff: Information is current as of January 2025.

Recommendations

  • For most users: Utilize the gguf/pauper_llama3_q4km.gguf with LM Studio for a balanced experience.
  • For maximum quality: Use the full HuggingFace model with the transformers library.
  • For low VRAM environments: The Q4_K_M quantization is suitable, requiring approximately 5GB.

Finetunes
1 models