burner22/numbat · Discussions

#45 opened about 10 hours ago by

kimi-chan

Report

#44 opened about 11 hours ago by

itsnotsplat

I’m a believer

➕👍 2

6

#43 opened about 12 hours ago by

fdsljwre

Make README.md tell true

#42 opened about 17 hours ago by

NotRootMint

Nice random number generated "weights"

#41 opened about 18 hours ago by

tsuchinoko-html

If You Fix It Like This, It Actually Looks Plausible — A Proposal on LatentMoE Projection and Depth Scaling for the 14T Numbat Architecture.

🔥 4

#39 opened about 23 hours ago by

Gavin-chen

It might be fake.

9

#38 opened about 23 hours ago by

mondk

Created for "engagement"

#37 opened about 23 hours ago by

MihaiPopa-1

One thing that keeps sticking in my head is not the alleged 7.7 TB size, but how clean the shard layout is.

#36 opened about 23 hours ago by

ChefThomas

One thing I'd love to hear more about from both believers and skeptics

#35 opened about 23 hours ago by

ChefThomas

This scam 🛑

#34 opened about 24 hours ago by

Xhub1880

pull deese nuts out your mouth

#33 opened about 24 hours ago by

bananabreads

Пишу в эпичном треде

🔥 1

#32 opened 1 day ago by

pdsg

Independent audit: 377/377 norm and 258/258 conv tensors reproduce bit-exactly from `sha256(tensor_name)`; 192M FP4 sign bits match 100%

❤️ 2

#31 opened 1 day ago by

ressl

Impresive! I runned it on my RTX 1050 Ti with quant UD-Q0_K_XL

#29 opened 1 day ago by

NotRootMint

🚩 Report: Spam

#28 opened 1 day ago by

cocorang

brb buying 7.0 TB of ram to test it out

#26 opened 1 day ago by

lastbattle

Wouldn't it be interesting to try training weights?

5

#25 opened 1 day ago by

testamentaddress01

我发现了破案的线索

😎🧠 11

#24 opened 1 day ago by

ACCC1380

🚩 Report: Illegal or restricted content

#23 opened 1 day ago by

cdnuts

🚩 Report: Spam

#22 opened 1 day ago by

fade-chan

🚩 Report: Spam

#21 opened 1 day ago by

Voidic98

EVAL: FAKE AF

#20 opened 1 day ago by

gbuzhf

Fake + larp

❤️ 8

#19 opened 2 days ago by

cekal

Cool prank bro.

#18 opened 2 days ago by

SigmaEsau

instagram author: eno30100

👍 1

#17 opened 2 days ago by

Exmanq

35b-a3b

🔥😎 29

#16 opened 2 days ago by

inikishev

my ass itches

❤️ 1

5

#15 opened 2 days ago by

bananabreads

Fake - Reason: 550B active -> too expensive + slow to serve

#14 opened 2 days ago by

shadowlilac

43 complete tensors exactly match name-seeded NumPy RNG output

🔥👍 22

6

#13 opened 2 days ago by

FloxUnder

🚩 Report: Spam

#12 opened 2 days ago by

iQuick

It is FAKE

8

#11 opened 2 days ago by

mdrbal83

🚩 Report: Illegal or restricted content

#10 opened 2 days ago by

ichugwindex

It's better to upload to the torrent

#9 opened 2 days ago by

Manlit

some question and concerns

#8 opened 2 days ago by

rairenter

Anthropic might have ulterior motives

5

#7 opened 2 days ago by

kubika

Beware of this model, Hugging Face automatically reports 6.8T parameters, while the uploader claims approximately 14T.

9

#6 opened 2 days ago by

Metris

Nice leak, let me spin up my spare 8xGB300 cluster

#5 opened 2 days ago by

sbcodes

Finally a small-size open-source model for my local GPU 😌

❤️ 13

#4 opened 2 days ago by

Vxtzq

Did an agent upload this?

👀 7

#3 opened 2 days ago by

Yules

🚩 Report: Illegal or restricted content

#2 opened 2 days ago by

euphoma

Wheres the rest