---
license: gemma
library_name: gguf
base_model: 0xSero/gemma-4-21b-a4b-it-REAP
tags:
  - gemma
  - gemma-4
  - moe
  - pruning
  - reap
  - gguf
  - llama-cpp
language:
  - en
pipeline_tag: text-generation
---

# Gemma-4 21B-A4B-it REAP - GGUF

GGUF quantized versions of [0xSero/gemma-4-21b-a4b-it-REAP](https://huggingface.co/0xSero/gemma-4-21b-a4b-it-REAP).

## Model Description

This is **20% expert-pruned** version of Google's Gemma-4 26B-A4B-it using **[Cerebras REAP](https://github.com/cerebras/reap)** (Router-weighted Expert Activation Pruning).

### Key Specifications

| Metric | Original (26B) | This Model (21B) |
|--------|----------------|------------------|
| Total params | ~26B | **21.34B** |
| Experts/layer | 128 | **103** |
| Active params/token | ~4B | ~4B |
| Disk size | ~52GB | **~43GB** |

REAP removes 20% of MoE experts (25 of 128 per layer) while preserving the model's routing behavior. The active parameter count per token is unchanged since the router still selects 8 experts per token from the remaining pool.

### Architecture
- **30 transformer layers**
- **Sliding attention** (window=1024) for 25 layers, **full attention** every 6th layer
- **MoE FFN** with 103 experts per layer, 8 active per token
- **Thinking model** -- uses `<|channel>thought` / `<|channel>response` channels
- **Multimodal** -- supports text and vision inputs
- **Context window:** 262,144 tokens

## Available Quantizations

| Filename | Quant Type | Size | Description |
|----------|------------|------|-------------|
| `gemma-4-21b-a4b-it-REAP.gguf` | BF16 | ~43GB | Full precision, best quality |
| `gemma-4-21b-a4b-it-REAP-Q8_0.gguf` | Q8_0 | ~23GB | High quality |
| `gemma-4-21b-a4b-it-REAP-Q5_K_M.gguf` | Q5_K_M | ~15GB | Balanced (recommended) |
| `gemma-4-21b-a4b-it-REAP-Q4_K_M.gguf` | Q4_K_M | ~13GB | Good quality, smaller |
| `gemma-4-21b-a4b-it-REAP-Q3_K_M.gguf` | Q3_K_M | ~10GB | Smallest |

## Usage with llama.cpp

```bash
# Download a quantized model
wget https://huggingface.co/Ayodele01/gemma-4-21b-a4b-it-REAP-GGUF/resolve/main/gemma-4-21b-a4b-it-REAP-Q5_K_M.gguf

# Run with llama.cpp
./llama-cli -m gemma-4-21b-a4b-it-REAP-Q5_K_M.gguf \
  -p "Write a quicksort in Python." \
  -n 2048
```

## Usage with Ollama

Create a Modelfile:
```
FROM ./gemma-4-21b-a4b-it-REAP-Q5_K_M.gguf

TEMPLATE """<bos><start_of_turn>user
{{ .Prompt }}<end_of_turn>
<start_of_turn>model
"""

PARAMETER stop "<end_of_turn>"
PARAMETER temperature 0.7
```

Then:
```bash
ollama create gemma4-21b-reap -f Modelfile
ollama run gemma4-21b-reap
```

## Benchmark Results (from original REAP model)

| Task | Original (26B) | REAP 21B |
|------|----------------|----------|
| Elementary Math | 92% | 90% |
| Philosophy | 92% | 88% |
| GSM8K | 86% | 84% |

Generation quality is "essentially indistinguishable from the original" according to the REAP authors.

## License

This model is released under the [Gemma License](https://ai.google.dev/gemma/terms).

## Credits

- **Original model:** [google/gemma-4-26b-a4b-it](https://huggingface.co/google/gemma-4-26b-a4b-it)
- **REAP pruning:** [0xSero/gemma-4-21b-a4b-it-REAP](https://huggingface.co/0xSero/gemma-4-21b-a4b-it-REAP)
- **REAP paper:** [arxiv.org/abs/2510.13999](https://arxiv.org/abs/2510.13999)
- **GGUF conversion:** Ayodele01