Content
@
https://warpcast.com/~/channel/llm
0 reply
0 recast
0 reaction
Jim
@zkp
i still use Venice because it claims to be private but its almost unusable compared to other LLMs are there any other privacy-focused options?
2 replies
0 recast
3 reactions
J. Valeska 🦊🎩🫂
@jvaleska.eth
which model are you using? new large uncensored model (mistral) is good enough and very fast could be 2-3 months max diff with openai top model, meaning everyone was saying wowow and losing their minds last 2 months for similar quality models imo and, image models are very good now too, open source is a bit slow, lack of billions 😅
1 reply
0 recast
2 reactions
Jim
@zkp
i think im using the right one? its the pro version i dont need it to be particularly fast or performant but it spends so much time “thinking”, and the responses are irrelevant and uninformed about certain topics. the thinking monologue is not encouraging either as it seems to get confused with simple requests. this is just for chat btw. totally agree most of the ai hype every few weeks is marketing. i will say grok impresses me with meme comprehension and nuance (in my experience anyway). goes to show the value of training data. venice on the other hand, even though its uncensored in some ways, seems like its trained on wikipedia or other government-approved information.
1 reply
0 recast
1 reaction
J. Valeska 🦊🎩🫂
@jvaleska.eth
yeah, this one if you click the I for info you can see the model and research on training data (it is open source) I found that thinking a bit slow too but I think it is faster now (or I have reasoning disabled or something, since it comes back in 1-2 sec and it was 30 seconds in the past)
1 reply
0 recast
1 reaction
Jim
@zkp
thanks for the info! maybe i just have to tweak the settings a bit.
1 reply
0 recast
1 reaction
J. Valeska 🦊🎩🫂
@jvaleska.eth
they are also changing things often, that's early tech
0 reply
0 recast
1 reaction
