File size: 1,148 Bytes
af6ea74
 
 
7a2a7c6
 
250f83c
af6ea74
 
250f83c
af6ea74
809f093
af6ea74
809f093
af6ea74
 
 
 
d11c45d
 
809f093
 
 
af6ea74
 
809f093
 
af6ea74
 
 
 
 
d11c45d
 
af6ea74
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
---
license: apache-2.0
inference: false 
base_model: teknium/OpenHermes-2.5-Mistral-7B
base_model_relation: quantized 
tags: [green, p7, llmware-chat, ov]
---

# teknium-open-hermes-2.5-mistral-ov

**teknium-open-hermes-2.5-mistral-ov** is an OpenVino int4 quantized version of teknium's popular open hermes finetune of mistral, providing a very fast, very small inference implementation, optimized for AI PCs using Intel GPU, CPU and NPU.    

[**teknium-open-hermes-2.5-mistral**](https://huggingface.co/teknium/OpenHermes-2.5-Mistral-7B) is a leading chat finetuned version of mistral 7b.  


### Model Description

- **Developed by:** teknium
- **Quantized by:** llmware   
- **Model type:** mistral-7b  
- **Parameters:** 7 billion
- **Model Parent:** teknium/OpenHermes-2.5-Mistral-7B
- **Language(s) (NLP):** English  
- **License:** Apache 2.0  
- **Uses:** General purpose chat
- **RAG Benchmark Accuracy Score:** NA  
- **Quantization:** int4  
  

## Model Card Contact

[llmware on github](https://www.github.com/llmware-ai/llmware)  

[llmware on hf](https://www.huggingface.co/llmware)  

[llmware website](https://www.llmware.ai)