Glm 5.2 reap

This is wild.

I wonder how it would perform with my EPYC, I’ve got 4 SSDs I can RAID0 for this sort of tomfoolery.

Will report back when I get the time.

Ah, someone beat me to it.

Was messing around with Colibri yesterday, pretty wild that it even worked at all.

Got like 0.6 t/s with my 285K, 64GB of RAM and MP700 Gen 5 NVME.

Too bad it doesn’t seem to properly implement the OpenAI API when I tried to intergrate it with opencode

1 Like

I’d read the article first because the video gives extra commentary on top of the article


Holy hell 330 tokens/second on the dspark deepseek v4 flash
If you have only time to consume 1, just watch the video

@wendell @eousphoros are you guys going to implement dspark on kappa?

1 Like

Nice. This video reminds me of how I judged my cpu upgrades back in the day. Pick a directory with lots of stuff in it and do a dir (DOS) to see how fast the directory gets listed.