shiningfractal (![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png)
shiningfractal) wrote2025-04-09 05:11 pm
shiningfractal) wrote2025-04-09 05:11 pm
(no subject)
hey army bros
try some open source chat for your
Sergeant Max
give it a prompt to roleplay and never break character
and then if you know how train it with all your institutional data
shouldn't take too long
No, ChatGPT's code is *not* open source.
### Details:
- The core models behind ChatGPT (like GPT-3.5 or GPT-4) are proprietary — developed by OpenAI.
- OpenAI has published research papers explaining *some* of the architecture and methods, but the actual code and training data are closed.
- The API and ChatGPT product (like you're using now) are available as paid services.
---
### What *is* open source?
However, there *are* open-source alternatives inspired by ChatGPT:
| Project | Model | Notes |
|--------|-------|-------|
| GPT-J / GPT-NeoX | EleutherAI | Open-source GPT-3-style models |
| LLaMA | Meta | Released for research use — widely adapted |
| Mistral | Mistral AI | Open-weight models performing very well |
| OpenChat | Community | Fine-tuned open models for chat use |
| Ollama | Framework | Easy local deployment of chat models |
| Hugging Face | Various | Huge collection of open-source models |
---
### Can I run "ChatGPT" locally?
Not exactly — but you can run good open models that are *close* in performance, like:
- Mistral 7B / Mixtral 8x7B
- LLaMA 2 / LLaMA 3
- OpenChat / Zephyr
- Phi-2 (Microsoft) for small devices
- GPT-J / GPT-NeoX
> Tools like *Ollama* or *LM Studio* make this very easy.
---
Want me to help you set one up locally? Or compare models for a specific purpose (e.g., coding, chat, roleplay, research)?
***
try some open source chat for your
Sergeant Max
give it a prompt to roleplay and never break character
and then if you know how train it with all your institutional data
shouldn't take too long
No, ChatGPT's code is *not* open source.
### Details:
- The core models behind ChatGPT (like GPT-3.5 or GPT-4) are proprietary — developed by OpenAI.
- OpenAI has published research papers explaining *some* of the architecture and methods, but the actual code and training data are closed.
- The API and ChatGPT product (like you're using now) are available as paid services.
---
### What *is* open source?
However, there *are* open-source alternatives inspired by ChatGPT:
| Project | Model | Notes |
|--------|-------|-------|
| GPT-J / GPT-NeoX | EleutherAI | Open-source GPT-3-style models |
| LLaMA | Meta | Released for research use — widely adapted |
| Mistral | Mistral AI | Open-weight models performing very well |
| OpenChat | Community | Fine-tuned open models for chat use |
| Ollama | Framework | Easy local deployment of chat models |
| Hugging Face | Various | Huge collection of open-source models |
---
### Can I run "ChatGPT" locally?
Not exactly — but you can run good open models that are *close* in performance, like:
- Mistral 7B / Mixtral 8x7B
- LLaMA 2 / LLaMA 3
- OpenChat / Zephyr
- Phi-2 (Microsoft) for small devices
- GPT-J / GPT-NeoX
> Tools like *Ollama* or *LM Studio* make this very easy.
---
Want me to help you set one up locally? Or compare models for a specific purpose (e.g., coding, chat, roleplay, research)?
***