FrenzyMath/Reaper-Base
Reaper-Base by FrenzyMath is a 1.72 billion parameter base checkpoint utilizing the Qwen3 causal language model architecture with BF16 weights. Designed for research and as an initialization point for subsequent models in the Reaper family, it features a context length of 131,072 tokens and a vocabulary size of 151,936. This model is primarily intended for further fine-tuning and evaluation within specific domains.
Loading preview...
Overview
Reaper-Base is the foundational checkpoint within the Reaper model family, developed by FrenzyMath. It is built upon the Qwen3 causal language model architecture, featuring approximately 1.72 billion parameters and utilizing BF16 weights. This model serves as a crucial starting point for research and for initializing more advanced checkpoints within the Reaper collection.
Key Capabilities
- Base Checkpoint: Provides a robust foundation for further supervised fine-tuning (SFT) and policy-value checkpoint development within the Reaper family.
- Qwen3 Architecture: Leverages the capabilities of the Qwen3 causal language model.
- Large Context Window: Configured with a context length of 131,072 tokens, allowing for processing extensive inputs.
- Extensive Vocabulary: Features a vocabulary size of 151,936, supporting a broad range of linguistic expressions.
Good For
- Research and Development: Ideal for researchers and developers looking to experiment with and build upon a base language model.
- Initializing Fine-tuning: Specifically designed to be the starting point for supervised fine-tuning (SFT) and other training stages.
- Exploring Qwen3 Architecture: Useful for those interested in working with models based on the Qwen3 causal language model architecture.