Being that the 8B is Distilled on LLama, I wonder how the Qwen 7B version would do being that it's also a Chinese model. I imagine this is what's affecting the model. Maybe ask the model about Facebook/Meta, lol. P.S. Also, maybe enable "DeepThink (R1)" in the chat, maybe that will make the results closer to each other as this is the default of the Ollama models.
Most of the text in the terminal is the “thinking” portion before its “response”. In app form you can choose to hide the thinking text to just see the response if you like.
lol i cracked up when it said "WE firmly believe that under the grand cause..." . It sounds like something Big Brother from 1984 would say.
😅
Being that the 8B is Distilled on LLama, I wonder how the Qwen 7B version would do being that it's also a Chinese model. I imagine this is what's affecting the model. Maybe ask the model about Facebook/Meta, lol. P.S. Also, maybe enable "DeepThink (R1)" in the chat, maybe that will make the results closer to each other as this is the default of the Ollama models.
Thanks for the ideas!
Deepseek is extremely verbose, don't like it
Most of the text in the terminal is the “thinking” portion before its “response”. In app form you can choose to hide the thinking text to just see the response if you like.