Nick Daniel

Pelicans, Saints, & Miss St Fan ⚜️ La Tech Golf Alum ⛳️ Former Miss St Men’s Basketball Asst. Video Coordinator

[Nick Daniel @Nickdaniel30]

@ItsmeAjayKV @MiniMax_AI Doubled in price though right?

[Nick Daniel @Nickdaniel30]

@antirez Well said

[Nick Daniel @Nickdaniel30]

@dealignai 💪🏼

[Nick Daniel @Nickdaniel30]

@DJLougen Just joined 💪🏼

[Nick Daniel @Nickdaniel30]

@dealignai @jun_song Great stuff man

[Nick Daniel @Nickdaniel30]

@jun_song Going to be groundbreaking

[Nick Daniel @Nickdaniel30]

@ai_hakase_ Insane. What are you serving it on & all your flags/sampling etc? I haven’t been able to reach those numbers on my M1 Ultra 128gb!

[Nick Daniel @Nickdaniel30]

@ivanfioravanti AI is going to be insane by the end of the year

[Nick Daniel @Nickdaniel30]

@witcheer Yes let us know

[Nick Daniel @Nickdaniel30]

@antirez How are you serving it/runtime? I’m only getting 13 tok per sec on my M1 Ultra 128gb & it’s driving me nuts. Can’t fix it. I’m using JANGTQ2

[Nick Daniel @Nickdaniel30]

@dealignai Hey brother I’ve got a M1 Ultra 128GB & I can only run your DSV4 JANGTQ2 at like 13 tok/sec… I’ve tried everything. What runtime are you using & do you have any advice? I know it should be over 40+ tok/sec.

[Nick Daniel @Nickdaniel30]

@jun_song I’ve got a M1 Ultra 128GB & I can only run DSV4 Flash at like 13 tok/sec… I’ve tried everything. How are you serving yours & do you have any advice? I know it should be over 40+ tok/sec

[Nick Daniel @Nickdaniel30]

27B FP16 MTP with a 2.42x speed up on Hermes! 128GB M1 Ultra on llama.cpp @DJLougen

[Nick Daniel @Nickdaniel30]

@KyleHessling1 @NousResearch So fire

[Nick Daniel @Nickdaniel30]

@KyleHessling1 @NousResearch So are you saying for agentic use like in Hermes all of these qwen models are best with thinking on & temp @ 1.0?

[Nick Daniel @Nickdaniel30]

@jun_song I bought the M1 ultra 128gb instead of the M5 max… did I choose poorly? 😂

[Nick Daniel @Nickdaniel30]

@TeksEdge Also had crazy results with Q8 on my M1 Ultra 128GB

[Nick Daniel @Nickdaniel30]

@Youssofal_ Really ground breaking stuff man. Next step is matching llama.cpp’s consistency for agentic tool calls/stability. Would love to run this consistently on Hermes/Openclaw

[Nick Daniel @Nickdaniel30]

@AmarSVS Very interesting I don’t see the link in the reply though brother