---
library_name: transformers
pipeline_tag: text-generation
base_model: Qwen/Qwen3-1.7B-Base
tags:
- qwen3
- reaper
- causal-lm
---

# Reaper-Base

Reaper-Base is the base checkpoint released with the Reaper model family. It uses the Qwen3 causal language model architecture with approximately 1.72B parameters and BF16 weights.

This checkpoint is part of the [Reaper model collection](https://huggingface.co/collections/FrenzyMath/reaper-6ad03ad88957749d6114b8f6).

## Model details

- Architecture: `Qwen3ForCausalLM`
- Parameters: 1,720,574,976
- Weight format: `safetensors`
- Context length in the model configuration: 131,072 tokens
- Vocabulary size: 151,936

## Model family

| Checkpoint | Stage |
| --- | --- |
| [FrenzyMath/Reaper-Base](https://huggingface.co/FrenzyMath/Reaper-Base) | Base checkpoint |
| [FrenzyMath/Reaper-SFT](https://huggingface.co/FrenzyMath/Reaper-SFT) | Supervised fine-tuning |
| [FrenzyMath/Reaper-1](https://huggingface.co/FrenzyMath/Reaper-1) | Final policy-value checkpoint |

## Usage

```python
from transformers import AutoModelForCausalLM, AutoTokenizer

repo_id = "FrenzyMath/Reaper-Base"

tokenizer = AutoTokenizer.from_pretrained(repo_id)
model = AutoModelForCausalLM.from_pretrained(
    repo_id,
    torch_dtype="auto",
    device_map="auto",
)
```

Use a recent version of `transformers` with Qwen3 support.

## Intended use and limitations

This release is intended for research and for initializing later checkpoints in the Reaper family. Users should independently evaluate the model for their target domain. Training-data details, benchmark results, and a license declaration are not included in the current release.