---
license: gemma
tags:
  - gemma
  - gguf
  - llama.cpp
  - wllama
  - legal
  - auto-completion
---

# Gemma 3 270M - Legal Auto-Completion (GGUF)

Fine-tuned Gemma 3 270M for legal text auto-completion, optimized for client-side browser deployment.

## Model Details

- **Format**: GGUF (Q4_K_M quantization with Q5_0 fallback)
- **Size**: ~169 MB
- **Optimized for**: Wllama (WebAssembly) browser inference

## Usage with llama.cpp

```bash
./llama-cli -m gemma-270m-q4km-optimized.gguf -p "The court held that" -n 30
```

## Usage with Wllama (Browser)

```javascript
import { Wllama } from '@wllama/wllama';

const CONFIG_PATHS = {
    'single-thread/wllama.wasm': 'path/to/single-thread/wllama.wasm',
    'multi-thread/wllama.wasm': 'path/to/multi-thread/wllama.wasm',
};

const wllama = new Wllama(CONFIG_PATHS);
await wllama.loadModelFromHF(
    'Ashraya/gemma-270m-legal-gguf',
    'gemma-270m-q4km-optimized.gguf'
);

const result = await wllama.createCompletion('The court held that', {
    nPredict: 20,
    sampling: { temp: 0.3, top_p: 0.95 }
});
console.log(result);
```

## Files

- `gemma-270m-q4km-optimized.gguf` - Main model file (~169MB)