Running Kimi K3 on a M1 Max • Hacker News Board

Re: Running Kimi K3 on a M1 Max

#21

Post by embedding-shape » Tue, Jul 28, 2026, 10:20 PM UTC

tito wrote: post #7

Azantys wrote: post #3

0.01 tk/s is unusable for anything, you would wait a whole day for just 1000 token of output, what is the point of projects like this?

I like seeing the latest and greatest model crammed into new systems to see how it fares. To deal with the speed, one person on reddit suggested using it in an email interface rather than a chat interface.

Email would indeed be fitting for K3 running on a M1 Mac, as it'd take days/weeks to receive a response, which matches with my real-world emailing experience pretty well.

Re: Running Kimi K3 on a M1 Max

#22

Post by winstonp » Tue, Jul 28, 2026, 10:20 PM UTC

Azantys wrote: post #3

0.01 tk/s is unusable for anything, you would wait a whole day for just 1000 token of output, what is the point of projects like this?

16 tokens / s is not nothing.

Re: Running Kimi K3 on a M1 Max

#23

Post by tjwebbnorfolk » Tue, Jul 28, 2026, 10:21 PM UTC

> ~60–76 s/token

I don't know if I'd call this "running"

Re: Running Kimi K3 on a M1 Max

#24

Post by ggm » Tue, Jul 28, 2026, 10:23 PM UTC

Azantys wrote: post #3

0.01 tk/s is unusable for anything, you would wait a whole day for just 1000 token of output, what is the point of projects like this?

So you subscribe to the belief we won't in future find mentalism in other galaxies or solar systems which operate on mechanisms we don't understand and think v e r y s l o w w w w w w w l y ?

(note. I am not a believer in AGI)

"useful" is highly contextual. The clock of the long "now" is not useful in the sense you mean, to synchronise your wristwatch. I'm still glad it exists.

Re: Running Kimi K3 on a M1 Max

#25

Post by sermah » Tue, Jul 28, 2026, 10:23 PM UTC

tjwebbnorfolk wrote: post #23

> ~60–76 s/token I don't know if I'd call this "running"

Had the same feeling when I first saw min/km units in some (human) running context.

UPD: I know it's not the same at all, just the reversal of units that gets me

Re: Running Kimi K3 on a M1 Max

#26

Post by tito » Tue, Jul 28, 2026, 10:24 PM UTC

tjwebbnorfolk wrote: post #23

> ~60–76 s/token I don't know if I'd call this "running"

I commented similarly below, but as a terrible programmer, I probably perform about 1 minute per token too (at Kimi 3 level). It puts into context how I think about intelligence

Re: Running Kimi K3 on a M1 Max

#27

Post by mips_avatar » Tue, Jul 28, 2026, 10:26 PM UTC

Would be interesting to see how fast it would be on 4x mac studio 512gb machines.

Re: Running Kimi K3 on a M1 Max

#28

Post by brokencode » Tue, Jul 28, 2026, 10:27 PM UTC

tito wrote: post #26

tjwebbnorfolk wrote: post #23

> ~60–76 s/token I don't know if I'd call this "running"

I commented similarly below, but as a terrible programmer, I probably perform about 1 minute per token too (at Kimi 3 level). It puts into context how I think about intelligence

Maybe in terms of code produced, but one token is only a fragment of a thought for an LLM.

It’d be like thinking as slowly as Ents talk to each other in Lord of the Rings.

Re: Running Kimi K3 on a M1 Max

#29

Post by acmnrs » Tue, Jul 28, 2026, 10:29 PM UTC

The title should probably be edited to specify "M1 Max" instead of "M1 Mac". You aren't running K3 on a base M1 anytime soon. Either way, still a very impressive project.

Re: Running Kimi K3 on a M1 Max

#30

Post by tito » Tue, Jul 28, 2026, 10:30 PM UTC

acmnrs wrote: post #29

The title should probably be edited to specify "M1 Max" instead of "M1 Mac". You aren't running K3 on a base M1 anytime soon. Either way, still a very impressive project.

Done, Mac -> Max