Spaces:
Running
Why are you uploading fake models
Why are you uploading fake models
I am very confused, what justifies a "fake model"?
I am very confused, what justifies a "fake model"?
^^^
Why are you uploading fake models
They aren't. These are distillations of larger reasoning models, following a similar naming convention DeepSeek used eg:
deepseek-ai/DeepSeek-R1-Distill-Qwen-7B
^ That is Qwen2.5-7B trained on Deepseek-R1 traces
TeichAI have a much clearer naming convention
[Base-Model]-[Data-Source]-Distill-[number of samples]
eg: this is Qwen3-8B trained on 1000 samples from Gemini-3-Pro-Preview:
TeichAI/Qwen3-8B-Gemini-3-Pro-Preview-Distill-1000x
And they're even sharing the datasets they create:
https://huggingface.co/TeichAI/datasets
What are you contributing to the scene?
Yea I think he is being misled by the DeepSeek distills.
I think DeepSeek v3.2 was trained on chatgpt's responses at some point because the model seems to be convinced that it's chatgpt.
Either way the models are anything but fake. They are weak models trying to imitate the strong 🦾