Running Kimi K3 on a M1 Max

128

ttito about 14 hours ago 85 commentsRead Article on github.com

RU version is available. Content is displayed in original English for accuracy.

⚡ Community Insights

Discussion Sentiment

72% Positive

Analyzed from 1883 words in the discussion.

Discussion (85 Comments)Read Original on HackerNews

antirez•about 14 hours ago

SSD streaming on an M5 Max 128GB: https://x.com/antirez/status/2082136334160818528

Soon decent speed across two Mac Studios with 512GB of RAM.

tito•about 14 hours ago

wow cool! I like watching the new models come out and how they end up crammed in to run on local machines. I learned what mxfp4 is thanks to this latest Kimi release - although it sounds like it means that there's less room for compression in the model compared to others.

whatsThisBtn4•about 13 hours ago

0.3tx per second is decent speed?

And it gets worse with every token.

zozbot234•about 11 hours ago

That's not what parent said, but this is already quite decent speed for unattended inference (overnight or even spanning multiple business days) which is arguably the right target for this model. This is a challenging model to infer locally, it has roughly ~115 GB of dense active parameters(!) plus ~25 GB of sparsely routed experts per token. Plus the KV cache (which is actually reasonably lean for this one model, around 27GB for a full 1Mtok context). What you're seeing in antirez's video is essentially the performance we should expect from a 128GiB system that has to load sparsely routed experts in full from disk because there's no real room for caching them.

192GiB Gorgon Halo systems will be an interesting future target for this model, the best you can do with 128GiB or less is probably to push batching higher in order to amortize the weights traffic over multiple inferences - which of course will sink single-session speeds even lower for a modest gain in total throughput.

tito•about 9 hours ago

I read about its format being tough to compress further. What should we expect in terms of creating smaller models to run locally?

whatsThisBtn4•about 11 hours ago

Would it even finish over a few weeks?

For AI agents this would take a year

searealist•about 9 hours ago

No one will ever derive any utility from running models at this speed. Please prove me wrong. Give me the number of tokens input and output (and dont forget about reasoning) and acceptable time to wait for it and the use case.

smallerize•about 11 hours ago

That's not what he said. He said with 2x the hardware it will be faster.

xfour•about 13 hours ago

Cool stuff. Do you have a the hardware and a way to bridge the compute? Or just hopeful?

weikju•about 9 hours ago

Supposedly RDMA over Thunderbolt 5 can allow this.

api•about 12 hours ago

I’ve wondered for a while: given the lower cost of SSD per GB could you build a very wide RAID0 style striped array of SSDs (maybe one per slot) to get almost RAM like read speeds?

To really go fast you’d probably have to do PCB layout and do like 256 or 1024 chips in parallel with a fast SRAM aggregation buffer feeding a GPU or TPU rig.

Or could you do the same with custom layout of cheap slower RAM?

I wonder if anyone is doing this? You would flash in a model and then just run it. It would need RAM for context but much less of it.

colingauvin•about 7 hours ago

The problem is you'd have to traverse the PCI-E bus every transfer. Even with DMA it still has to physically get off the drive and onto the card. Even with the mythical PCI-E 6, if you had enough NVMEs to saturate, you have to do two transfers to get it to the inference hardware. And that tops out at 128 GB/s which is roughly the speed of DDR5 but with one extra hop.

Optane would actually be useful in this era. Intel was ahead of their time.

Azantys•about 14 hours ago

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?

SXX•about 14 hours ago

It is fun.

Also its answering the question of what gonna happen if you wake up tomorrow and datacenters are gone. Or internets are gone.

Some people on our globe live in countries with no internet whatsoever. Of course most of them dont have Macbook with 64GB RAM either, but it's much much easier to get than internet connection or rack of GB200.

SOTA LLMs are efficiently compression of all the knowkedge humanity has built. Having ability to run it at home to extract said knowledge is important no matter the speed.

fp64•about 8 hours ago

A post-apocalyptic desert landscape. The hero descends into the ruins, now a temple. Crude drawings of times before, people riding dinosaurs and mobile phones with arms and legs swinging a whip at the oppressed. He reaches the altar, a old MacBook, bottlecaps and fidgit spinners glued as a mosaic onto it. In the corner three starved men, riding on power generating bicycles providing some artificial light and supplying the machine with divine electricity. He proceeds and types: "we are starving. How can we get food?". Everybody waits in silence. Hours pass. Days. Slowly, letters emerge on the cracked screen: "you are absolutely right —" the screen flickers and fades. Panic. The link to God severed cynically. The hero has to seek out a new device, into distant lands. Maybe this time it will save them

tito•about 7 hours ago

“How long?” he said. “Seven and a half million years,” said Deep Thought.

whatsThisBtn4•about 13 hours ago

Even if somehow all the data centers are gone, it's still uselessly slow.

But also that's a pretty extreme hypothetical. Imagine the polymarket on that.

tito•about 14 hours ago

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.

magicalhippo•about 14 hours ago

> one person on reddit suggested using it in an email interface rather than a chat interface

Kimi Pen Pal. Bring back lettets and postcards. Do OCR, and use one of those 3D printer-like pen plotters write the model output as a letter.

Challenge would be automating the opening and OCR preparation, and the folding and mailing of the return letter. But given it's done commercially it should be possible.

embedding-shape•about 14 hours ago

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.

0xc133•about 11 hours ago

I had my clanker implement this idea in a standalone Rust server that speaks IMAP and SMTP and proxies your emails to an OpenAI endpoint you configure: https://tangled.org/clee.sh/posthorn

Works in mutt; other MUAs may vary.

tito•about 9 hours ago

lovely! I'm a fan of asynchronous LLM systems like this. There are times where I prefer slow, complete responses to immediate replies.

For example there's a billboard on 101 for Poppy Bank offering 4% interest and I asked my daily LLM to look into it through a voice note. The next morning, lo and behold it says it's an advertised rate, hard to actually get, and businesses aren't eligible. OK, done. Better than getting the response while I'm driving. And even if I wasn't driving, there's a level of, how do you say, it's easy to drop it the next day when I see it vs getting engrossed in the research.

hugopuybareau•about 14 hours ago

Love the email idea

tito•about 14 hours ago

Having the right type of interface makes a huge difference.

It reminds me of when Willow Garage chose to name their bot the TurtleBot, because if they named it anything else, people would think it was fast and capable. But when they called it Turtle Bot, people just kind of liked it and were satisfied with what it did.

At the level of Kimi 3, I probably can code only about 1,000 good tokens per day, too. (thankfully coding isn't my job)

ggm•about 14 hours ago

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.

tito•about 13 hours ago

Are there any well thought through stories about what this would look like? For example, I'm thinking about like nutrient flow, decision making, energy input, gravitational force, things like that seem to govern the value and speed of intelligence.

indiv0•about 12 hours ago

Hard to avoid spoilers here but Vernor Vinge hits on pretty much exactly this in A Fire Upon the Deep. Though he's interested less in the hard sci-fi aspects of how/why and more on the consequences of it (story-wise).

kirubakaran•about 13 hours ago

For the opposite, "Dragon's Egg" by Robert L. Forward is a fantastic read

ggm•about 12 hours ago

I think this is a slow version of the quandry behind Quantum Computing: how do you distinguish events from the noise floor? It happens in the quantum context and it would happen in the millenial timeframe completing "operations" which have to be compared to e.g. the stability of orbit around a sun.

whatsThisBtn4•about 13 hours ago

I can only guess some post purchase remorse.

Need to justify buying an expensive rig that doesn't do what you expected.

Specifically thinking the people they could do something AI with cpu, and realizing it isn't feasible. Happened at my fortune 20 company. They had to get approvals and ofc it was useless. Plenty people tried to explain, but they were the principle engineer, and out ranked everyone.

"It's not going to work", the topic changed, and we never spoke about it again.

anigbrowl•about 12 hours ago

'Large Language models? They can barely produce gibberish sentences, what would this tech ever be useful for?'

- bunch of people only ~4 years ago

Azantys•about 12 hours ago

I questioned the speed not the output quality, that is another discussion

bibstha•about 11 hours ago

It applies to speed too. The project paves way for more optimization at many layers overtime.

Francute•about 11 hours ago

You can't improve what you can't measure.

Consider this like if it were the first test

winstonp•about 14 hours ago

16 tokens / s is not nothing.

throwaway219450•about 13 hours ago

It's the other way around due to poor framing, it would be much easier to compare if you [the repo] said 0.02 tps.

whatsThisBtn4•about 13 hours ago

Confusing numbers everywhere.

16tk/s... Then 3 tks per minute. Then someone else posted 0.3tk/s.

Azantys•about 14 hours ago

? Readme says 60-70s per token

acmnrs•about 14 hours ago

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.

tito•about 14 hours ago

Done, Mac -> Max

mannyv•about 8 hours ago

Going to try this on my M1 Ultra 128gb.

The point of these engineering tricks is to see the envelope of what's possible. You can use these tricks to both run a bigger model on smaller hardware or run a smaller model on smaller hardware.

nlessard•about 14 hours ago

Anyone who knows the state of NVMe hardware more than me know if this would obliterate the lifespan of your drive? Seems like the biggest limitation to me (some people are probably fine with letting their Macs churn over the weekend).

trollbridge•about 13 hours ago

No problem at all to read data over and over. In fact, LLM weights are a great candidate for low-quality flash that can't handle a lot of write cycles, and you want a large amount of storage cheaply...

addaon•about 14 hours ago

Reads are not generally life-limiting for flash. (Well, no more so than power-on time in general. You still have aging mechanisms like electromigration, but these are orders of magnitude slower than write-induced damage.)

nlessard•about 14 hours ago

Cool, thanks

mips_avatar•about 14 hours ago

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

ALLTaken•about 13 hours ago

Exactly my machine 64GB M1 Max So happy about this! ♡

idk how people access (soldout) and even afford 512GB RAM MacStudio's. Isn't it $40k or so?

hmokiguess•about 13 hours ago

You're lucky, mine is the M1 Pro 16GB

teaearlgraycold•about 12 hours ago

I just checked eBay. There's an insane price difference between used and new. $5k vs $40k.

knighthacker•about 11 hours ago

Local AI is going to win. Not because it's cheaper btw.

jrhizor•about 12 hours ago

Super cool, and I appreciate the upfront speed disclaimer

mindwok•about 12 hours ago

Does anyone else feel like the writing is on the wall for a future of local models? Spamming data centres everywhere, powering them, having to commit insane capital to hardware, all the effort to serve inference over a network reliably - when here we are with a frontier model nearly running on a laptop.

Local AI on your device seems like a much more likely future to me than datacenters in space. For inference at least, training is another story.

nomel•about 12 hours ago

0.01 tk/s on an M1 Max is not "nearly". This is completely unusable, and in no way cost effective.

0.01 tokens per second means 1 million tokens ($3 worth of API usage [1]) takes 3.2 YEARS.

[1] https://www.kimi.com/resources/kimi-k3-pricing

mindwok•about 9 hours ago

Ok in terms of running a 2.8T parameter model, that's true.

Looking more broadly though, a model I can run on my laptop (Gemma 4) is ~4 points away from GPT-5.3 codex or Sonnet 4.5 on arena.ai LLM leaderboard. Those models were SOTA less than a year ago.

hmokiguess•about 13 hours ago

Now set it up with an agent and a permanent `/goal` to say it cannot stop until it has solved for speed, then leave it on and livestream so we can all see when it becomes exponential. Could have the Eternal Jukebox playing in the background!

Anoian•about 13 hours ago

That sounds like the most boring exciting livestream of all time.

denysvitali•about 13 hours ago

60s/token - if only there was a way to drop that "s" this would be amazing

als0•about 14 hours ago

Says it requires a 2TB disk? Must it be internal NVMe?

Fergusonb•about 14 hours ago

You can use an external drive if it's mounted as a writable volume. I would make sure it's fast, maybe thunderbolt 3/4/5 enclosure with a fast drive.

Xeoncross•about 13 hours ago

FYI, Thunderbolt 3 NVMe enclosures will be 10Gbps and 4 will be 40Gbps which is a big difference (and you'll notice it in the pricing as well)

tito•about 14 hours ago

The Github specifically mentions an option to stream it from an online host. It's extremely slow.

throwitaway222•about 11 hours ago

how much money in hardware would it cost to get 100tps?

tjwebbnorfolk•about 14 hours ago

> ~60–76 s/token

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

sermah•about 14 hours ago

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

tito•about 14 hours ago

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

brokencode•about 14 hours ago

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.

tito•about 13 hours ago

Oh, is that how it works? So, when somebody says a model is running at X tokens per second, it means that the thinking process is running at that, and output tokens are much lower then? Thanks to the explanation.

lostmsu•about 14 hours ago

under 0.02 tok/s

walrus01•about 12 hours ago

tokens/second, no, more like seconds/token

ncallaway•about 12 hours ago

The link beat you to it:

> It is not fast — about 16 seconds per token on our M1 Max

walrus01•about 11 hours ago

At least it's not minutes/token?

piterrro•about 14 hours ago

Will it fit on ESP32??

embedding-shape•about 14 hours ago

Better questions, how many ESP32s would it take to reach 1 tok/s decoding speed with K3?

tito•about 14 hours ago

1 button Kimi morse code interface

onesandofgrain•about 13 hours ago

Cool gimmick

brcmthrowaway•about 13 hours ago

Why not train another smaller LLM to give the same answers as Kimi K3?

dkarras•about 11 hours ago

why not zip the entire internet to 1MB so everyone can have a copy? because it is not possible - we don't know if it is possible. I mean we know it is impossible, but we don't know if it is possible to do it with acceptable quality loss.