Alelcv27/Llama3.2-3B-Breadcrumbs-Base-INST

TEXT GENERATIONConcurrency Cost:1Model Size:3.2BQuant:BF16Ctx Length:32kTool Calling:SupportedPublished:Apr 29, 2026Architecture:Transformer Cold

Alelcv27/Llama3.2-3B-Breadcrumbs-Base-INST is a 3.2 billion parameter language model created by Alelcv27, based on the Llama 3.2 architecture. This model is a merge of Llama-3.2-3B and Llama-3.2-3B-Instruct using the Model Breadcrumbs method, designed to combine the strengths of both base and instruction-tuned models. It features a notable context length of 32768 tokens, making it suitable for tasks requiring extensive contextual understanding.

Loading preview...

Model Overview

Alelcv27/Llama3.2-3B-Breadcrumbs-Base-INST is a 3.2 billion parameter language model built upon the Llama 3.2 architecture. It was created by Alelcv27 using the mergekit tool, specifically employing the Model Breadcrumbs merge method. This approach combines a base model with an instruction-tuned variant to potentially leverage the general knowledge of the base model and the conversational capabilities of the instruction-tuned model.

Key Capabilities

  • Merged Architecture: Integrates meta-llama/Llama-3.2-3B and meta-llama/Llama-3.2-3B-Instruct to create a hybrid model.
  • Breadcrumbs Merge Method: Utilizes a specific merging technique detailed in the Model Breadcrumbs paper, which involves layer-wise weighting.
  • Extended Context Window: Supports a substantial context length of 32768 tokens, beneficial for processing longer inputs and maintaining coherence over extended conversations or documents.

Good For

  • Applications requiring a balance between foundational language understanding and instruction-following capabilities.
  • Use cases where a 3.2 billion parameter model with a large context window is advantageous for efficiency and performance.
  • Developers interested in exploring models created with advanced merging techniques like Model Breadcrumbs.