47 comments

[ 3.6 ms ] story [ 116 ms ] thread
Does it not use Metal, on macOS? Would it be faster if it did?
(comment deleted)
That README hits all my “this is authored by an LLM” instincts. I presume the codebase is also written by an LLM?
I prefer to have such detailed readme's created by a LLM while iterating over no iteration documentation at all and usually the later is the standard.
Where can this 1tb k3.waste be downloaded?
Interesting project. The headline number (29 GB of RAM) is for 4k context.

From what I've read elsewhere, Kimi K3 is quite verbose in its thinking. At the quoted rate, it would generate only a total of 1.8k tokens in 1 hour. Is that enough for it to get any thinking done and produce output on more complicated prompts?

So if this Mac uses 30-50W, that's 40-60 tok/Wh...vs maybe 80k for a modern GPU cluster? So that's about 1000-2000x more power for the SSD streaming, unfortunately.
This sounds a lot like what the colibri project did for GLM-5.2. I'm a fan so keep at it!

justvugg.github.io/colibri

Neat! But, what do you do with a 0.5tk/s LLM?

Have you tried running it via llamacpp or other software that supports naive SSD offloading to compare speeds?

You get 8 nvmes set them up in raid 0/1 across two full pcie5x16 ports and you could reach up to 4ish tokens per second, presumably.
> Neat! But, what do you do with a 0.5tk/s LLM?

Hopefully resolve incidents faster without people pasting slop into the incident thread.

Approximate calculation is putting the cost at ~$5 per million tokens (assuming 42W sustained, 20¢/kWh), and that's excluding hardware and other costs.
How does this project compare to https://github.com/gavamedia/deltafin ?
Hi! I'm one of the deltafin devs. The biggest difference is that this is not actually a 100% "pure" uncut Kimi K3. This is requantized to 3-bit residual, whereas deltafin is the full real unaltered k3, through and through.

WASTE reads about 17 GB/token versus Deltafin’s 25.8 GB/pass—roughly. That's 34% less expert traffic, and some could argue a 34% reduction in quality.

Dear creator: you didn't ship the first draft of your code - why did you ship the first draft of your README??
I couldnt find anything explaining the name of this company on their website but is it okay that they’re riding on the name of an open source tool?

SQLite code itself is public domain but I’m not sure about the name.

It's too bad Optane PMem is dead
I don't get the fascination how is optane better than a modern nvme? I don't think it was faster.
Ok, this one will take just 30h (compared to that other project that would take 6.25 days) to start writing output tokens after you say hi in Claude Code.
looks helpful

your readme is overly verbose

agents don't need that and its extremely low signal for humans too

tell your language model to get it to the point

Be advised that the firm behind it ("sqliteai") had a nasty history of using non-open source licenses, e.g. Elastic License. I advise against using anything by them for this reason even if this project currently has an open license.
Claude might as well be .5 tok/s. I end up waiting several minutes and what it tells me could usually be summarized in under 100 words.

So I could potentially live with this if it was concise.

Standard llama.cpp can mmap the gguf, so it'll stay on disk if it doesn't fit on memory, and the kernel page cache will ensure the hot parts ("resident trunk") stay resident.

What's the benefit of a custom implementation at all?

are they allowed to use the "SQLite" name?
They say it's a waste that you pay for the tokens and then the inference provider pays for the electricity. Isn't that how everything works? I pay cucumbers and the farmers have to pay for the water and the fertilizer...

I hope that reasoning is an after-the-fact justification by the LLM that wrote this.

It's a ver interesting idea and I wouldn't mind trying it out, but with a smaller model. At 0.5t/s and reading many gigabytes from the SDD every second... I wonder if this wouldn't be extremely practical if targeting a 500gib or 250gib model, something that is still outside most consumers' laptop.

Pretty soon we'll have 3T param models down to 1 bit. They'll be able to tell you whether they're off or on.