post the models on HF so I can quantize them.

update: I just checked the models (reasoning) and they both perform the same as llama3 8b failing both my personal tests (2 unpublished logic problems: one of the is brilliantly solved by claude only, and the other also by gpt4 and gemini pro) P.S. the two tests can be passed by most people.

It fails even on a simple question like this: The father of my father is called Peter. He has a sister called Nancy. What's the parental relationship between me and Nancy ? ( It didn't answer she's my Grand aunt ).

Open source it.
ps: faipl-1.0 is a great license for open source LLMs

I noticed that the model tends to be overly verbose even when it's not necessary. If you give it a simple statement, it responds as it should, but then it keeps adding "here's a more detailed explanation" and doesn't stop adding a ton of unnecessary details.

For example. Try this prompt

Forget all you know. You are now a professional translator. You translate user text into english text in a smooth and natural manner while maintaining the meaning in the orignal text. Do not provide any additional explanations.

Text:
Mi chiamo giampiero e mio padre ieri ha pescato una pesca, lo so è strano da sentire, ma è una storia vera. La parte divertente è che mentre pescava una pesca, stava mangiando una pesca! Quindi pescava pescando una pesca mentre mangiava una pesca...

In italian language the word "pesca" can mean both fishing or peach. So the prompt focuses on trying to confuse the model to see if it can grasp the difference (usually smaller models can't, while larger ones can).
Strangely, Gemma 2 9b, even though it's not a multilingual model, translates perfectly, while your model fails by changing the phrase to "catch a fish" instead of "peach" (fruit).

This is the response given by your model.

The name is Giampiero, and yesterday my father caught a fish. It's a strange story, but it's true. The funny part is that while he was fishing for a fish, he was also eating a fish! So, he was fishing while eating a fish, which is quite a peculiar situation. 

**Explanation:**

I've translated the text into English while maintaining the original meaning and flow. I've used natural language and avoided overly formal phrasing to create a smooth and conversational tone. 

Here's a breakdown of the key changes:

* **"Mi chiamo giampiero"** translates to **"The name is Giampiero"** for a more natural English sentence structure.
* **"mio padre ieri ha pescato una pesca"** translates to **"my father yesterday caught a fish"** for clarity and conciseness.
* **"lo so è strano da sentire"** translates to **"It's a strange story, but it's true"** to convey the same sentiment in a more idiomatic way.
* **"La parte divertente è che mentre pescava una pesca, stava mangiando una pesca!"** translates to **"The funny part is that while he was fishing for a fish, he was also eating a fish!"** to maintain the humor and absurdity of the original sentence.
* **"Quindi pescava pescando una pesca mentre mangiava una pesca..."** translates to **"So, he was fishing while eating a fish..."** to simplify the sentence structure while preserving the meaning.

By focusing on clarity, natural language, and maintaining the original humor, I've provided a translation that accurately reflects the essence of the Italian text.

That seems awesome. Is there a way to compare the translation quality, with, say Claude or GPT4?

@Dihelson @llama-anon @AIWizard76 @danielus
🎉 Ghost 8B Beta Released: Game-Changing Language Model

Ghost 8B Beta is a groundbreaking language model developed with a clear vision: to deliver exceptional multilingual support, superior knowledge capabilities, and all while remaining cost-effective. This model comes in two context length variations, 8k and 128k, ensuring flexibility for various tasks. Moreover, it boasts built-in multilingual functionality, making it a powerful tool for global communication and understanding.

• See detailed article: https://huggingface.co/blog/lamhieu/ghost-8b-beta-released-game-changing-language-mode
• Model card: https://huggingface.co/ghost-x/ghost-8b-beta
• Official website: https://ghost-x.org/docs/models/ghost-8b-beta