---
language:
- en
library_name: transformers
license: apache-2.0
tags:
- gpt
- llm
- large language model
- h2o-llmstudio
thumbnail: >-
  https://h2o.ai/etc.clientlibs/h2o/clientlibs/clientlib-site/resources/images/favicon.ico
pipeline_tag: text-generation
quantized_by: h2oai
---

# h2o-danube2-1.8b-chat-GGUF
- Model creator: [H2O.ai](https://huggingface.co/h2oai)
- Original model: [h2o-danube2-1.8b-chat](https://huggingface.co/h2oai/h2o-danube2-1.8b-chat)

## Description

This repo contains GGUF format model files for [h2o-danube2-1.8b-chat](https://huggingface.co/h2oai/h2o-danube2-1.8b-chat) quantized using [llama.cpp](https://github.com/ggerganov/llama.cpp/) framework.

Table below summarizes different quantized versions of [h2o-danube2-1.8b-chat](https://huggingface.co/h2oai/h2o-danube2-1.8b-chat). It shows the trade-off between size, speed and quality of the models.


| Name                             | Quant method                      | Model size | MT-Bench AVG | Perplexity | Tokens per second |
|:----------------------------------|:----------------------------------:|:----------:|:------------:|:------------:|:-------------------:|
| [h2o-danube2-1.8b-chat-F16.gguf](https://huggingface.co/h2oai/h2o-danube2-1.8b-chat-GGUF/blob/main/h2o-danube2-1.8b-chat-F16.gguf)    | F16                               |  3.66 GB   |     5.60     | 8.02       | 797               |
| [h2o-danube2-1.8b-chat-Q8_0.gguf](https://huggingface.co/h2oai/h2o-danube2-1.8b-chat-GGUF/blob/main/h2o-danube2-1.8b-chat-Q8_0.gguf)   | Q8_0                              |  1.95 GB   |     5.51     | 8.02       | 1156              |
| [h2o-danube2-1.8b-chat-Q6_K.gguf](https://huggingface.co/h2oai/h2o-danube2-1.8b-chat-GGUF/blob/main/h2o-danube2-1.8b-chat-Q6_K.gguf)   | Q6_K                              |  1.50 GB   |     5.51     | 8.03       | 1131              |
| [h2o-danube2-1.8b-chat-Q5_K_M.gguf](https://huggingface.co/h2oai/h2o-danube2-1.8b-chat-GGUF/blob/main/h2o-danube2-1.8b-chat-Q5_K_M.gguf) | Q5_K_M                            |  1.30 GB   |     5.56     | 8.10       | 1172              |
| [h2o-danube2-1.8b-chat-Q5_K_S.gguf](https://huggingface.co/h2oai/h2o-danube2-1.8b-chat-GGUF/blob/main/h2o-danube2-1.8b-chat-Q5_K_S.gguf) | Q5_K_S                            |  1.27 GB   |     5.49     | 8.12       | 1107              |
| [h2o-danube2-1.8b-chat-Q4_K_M.gguf](https://huggingface.co/h2oai/h2o-danube2-1.8b-chat-GGUF/blob/main/h2o-danube2-1.8b-chat-Q4_K_M.gguf) | Q4_K_M |  1.11 GB   |     5.60     | 8.27       | 1162              |
| [h2o-danube2-1.8b-chat-Q4_K_S.gguf](https://huggingface.co/h2oai/h2o-danube2-1.8b-chat-GGUF/blob/main/h2o-danube2-1.8b-chat-Q4_K_S.gguf) | Q4_K_S |  1.06 GB   |     5.59     | 8.34       | 1270              |
| [h2o-danube2-1.8b-chat-Q3_K_L.gguf](https://huggingface.co/h2oai/h2o-danube2-1.8b-chat-GGUF/blob/main/h2o-danube2-1.8b-chat-Q3_K_L.gguf) | Q3_K_L |  0.98 GB   |     5.23     | 8.72       | 1442              |
| [h2o-danube2-1.8b-chat-Q3_K_M.gguf](https://huggingface.co/h2oai/h2o-danube2-1.8b-chat-GGUF/blob/main/h2o-danube2-1.8b-chat-Q3_K_M.gguf) | Q3_K_M |  0.91 GB   |     4.91     | 8.81       | 1107              |
| [h2o-danube2-1.8b-chat-Q3_K_S.gguf](https://huggingface.co/h2oai/h2o-danube2-1.8b-chat-GGUF/blob/main/h2o-danube2-1.8b-chat-Q3_K_S.gguf) | Q3_K_S |  0.82 GB   |     4.03     | 10.12      | 1103              |
| [h2o-danube2-1.8b-chat-Q2_K.gguf](https://huggingface.co/h2oai/h2o-danube2-1.8b-chat-GGUF/blob/main/h2o-danube2-1.8b-chat-Q2_K.gguf)   | Q2_K |  0.71 GB   |     3.03     | 12.56      | 1160              |

Columns in the table are:
* Name -- model name and link
* Quant method -- quantization method
* Model size -- size of the model in gigabytes
* MT-Bench AVG -- [MT-Bench](https://arxiv.org/abs/2306.05685) benchmark score. The score is from 1 to 10, the higher, the better
* Perplexity -- perplexity metric on WikiText-2 dataset. It's reported in a perplexity test from llama.cpp. The lower, the better
* Tokens per second -- generation speed in tokens per second, as reported in a perplexity test from llama.cpp. The higher, the better. Speed tests are done on a single H100 GPU


## Prompt template
```
<|prompt|>Why is drinking water so healthy?</s><|answer|>
```