BTW, these are the #AI #LLM models I settled on using with #JanAI:
#Qwen2.5 at 0.5B (Qwen2_5-0_5B-Instruct-uncensored_Q8_0), for fastest performance on low-end hardware
#Qwen2 at 1.5B (Qwen2-1_5B-Instruct-Abliterated-Q5_K_M), for balanced performance and good enough output quality
#Llama3.2 at 3B (Llama-3_2-3B-Instruct-heretic-ablitered-uncensored_Q5_K_M), for higher quality output
#Llama3 actually doesn’t run too poorly on my machine, although it can take some time to load up responses sometimes.








