---
base_model: miniHui/Geo-R1
pipeline_tag: image-text-to-text
library_name: gguf
license: mit
tags:
- gguf
- qwen2.5-vl
- multimodal
- quantized
---

# Geo-R1 GGUF

**This repo is proudly made by TerraByte AI.**

GGUF conversion of [miniHui/Geo-R1](https://huggingface.co/miniHui/Geo-R1), a geospatial reasoning model based on Qwen2.5-VL-7B-Instruct.

## Files

| File | Format | Size | SHA-256 |
| --- | --- | ---: | --- |
| `Geo-R1-Q4_K_M.gguf` | Q4_K_M language model (4.91 BPW) | 4,683,072,352 bytes | `666dc005f95b5cdbc2a8cb578621fac0ef71391b6b4f688d75422201ce5756ba` |
| `mmproj-Geo-R1-F16.gguf` | F16 vision projector | 1,354,163,200 bytes | `8dd5314810acfd740d0e75da8ceb19638d0861dd17107336c551711751ad7303` |

The language model is quantized to 4-bit `Q4_K_M`. The vision projector is kept at F16 to preserve image understanding quality and is required for image or video inputs.

## Usage

Use a recent [llama.cpp](https://github.com/ggml-org/llama.cpp) build:

```bash
llama-cli \
  -hf miniHui/Geo-R1-GGUF:Q4_K_M \
  --image /path/to/image.jpg \
  -p "Analyze this image and explain your reasoning."
```

To use local files explicitly:

```bash
llama-cli \
  -m Geo-R1-Q4_K_M.gguf \
  --mmproj mmproj-Geo-R1-F16.gguf \
  --image /path/to/image.jpg \
  -p "Analyze this image and explain your reasoning."
```

## Conversion

Converted and tested with llama.cpp revision `86a9c79f866799eb0e7e89c03578ccfbcc5d808e`.

The Q4_K_M model was validated with text generation, and the model/projector pair was validated with an image prompt using `llama-cli`.

See the [original model card](https://huggingface.co/miniHui/Geo-R1) for model details, intended use, and attribution.