AgentCPM-Explore GGUF Model

This skill provides access to the quantized GGUF model files for AgentCPM-Explore, a conversational AI model optimized for agent-based tasks. The model is available in multiple quantization levels to balance quality, speed, and memory usage.

Model Overview

Available Quantizations

The model is available in 25+ quantization variants, ranging from highly compressed (IQ1_S at 1.3GB) to near-full quality (Q6_K at 3.7GB). Each quantization offers different tradeoffs:

Recommended Variants

Low Resource Options

Extreme Compression

Usage Instructions

1. **Download the Model**

- Visit the model repository at `https://huggingface.co/mradermacher/AgentCPM-Explore-i1-GGUF`

- Select a quantization variant based on your requirements

- Download the corresponding `.gguf` file

2. **Load in Your Runtime**

For llama.cpp:

```bash

./main -m AgentCPM-Explore.i1-Q4_K_M.gguf -p "Your prompt here"

```

For Ollama:

```bash

ollama create agentcpm -f Modelfile

ollama run agentcpm

```

For LM Studio or Jan: Import the GGUF file through the UI.

3. **Configure Parameters**

- Set context length based on your use case

- Adjust temperature (0.7-0.9 recommended for conversational tasks)

- Configure top-p and top-k sampling as needed

4. **Multi-Part Files**

- If using split files, concatenate them before loading:

```bash

cat file-part1 file-part2 > complete-model.gguf

```

Quantization Selection Guide

When choosing a quantization:

1. **Available Memory**: Choose the largest quantization that fits your VRAM/RAM

2. **Speed Requirements**: Lower quantizations (Q4, IQ3) are faster

3. **Quality Needs**: Higher quantizations (Q5, Q6) preserve more model capability

4. **Use Case**: Conversational tasks benefit from Q4_K_M or higher

Model Capabilities

This model is designed for:

Important Notes

Constraints

Credits

Quantized by mradermacher with compute resources provided by nethype GmbH and @nicoboss. Original model by OpenBMB.