Great setup, how did you manage to connect them to cluster? What speeds do you get on 70b non moe models like KAT-DEV72B Q8 with context window 0k and 128k ?

Thanks.

New activity in Kwaipilot/KAT-Dev-72B-Exp 4 months ago

the instruct following is poor

#9 opened 4 months ago by

pypry

New activity in Kwaipilot/KAT-Dev-FP8 4 months ago

How much VRAM is required for this one

#1 opened 4 months ago by deleted

New activity in Kwaipilot/KAT-Dev 4 months ago

KAT-DEV 33B no longer working with Jan.ai 0.7.2 - Cline / LMstudio v0.3.31b4 - Cline

#12 opened 4 months ago by

akierum

Is KAT-Dev supposed to be that much slower than QWEN?

#11 opened 4 months ago by deleted

New activity in Mungert/KAT-Dev-72B-Exp-GGUF 4 months ago

KAT-DEV Q4 - Jan.ai - Roocode / Cline does not work Lmstudio crashes too (SOLVED)

#1 opened 4 months ago by

akierum

New activity in YOYO-AI/Qwen3-30B-A3B-YOYO-V4 4 months ago

coder vs yoyo-v4 vs thinking vs instruct in compex programming task

#2 opened 5 months ago by

NIK2703

New activity in unsloth/README 4 months ago

Request GGUF: Kwaipilot/KAT-Dev (62.4% (!) on SWE-Bench Verified having just 32B)

#15 opened 4 months ago by

Reverger

New activity in Mungert/KAT-Dev-GGUF 4 months ago

There is new KAT-Dev-72B-Exp

#1 opened 5 months ago by

akierum

New activity in mradermacher/KAT-Dev-GGUF 4 months ago

This model gives errors in cline, but new official one does not

👀 1

#3 opened 5 months ago by

akierum

New activity in Kwaipilot/KAT-Dev 4 months ago

why Kat-Dev worse then Qwen3-coder 30B?

#8 opened 5 months ago by

akierum

New activity in Mungert/Qwen3-Coder-30B-A3B-Instruct-GGUF 5 months ago

errors out with roocode

#1 opened 5 months ago by

akierum

JM

AI & ML interests

Recent Activity

Organizations

akierum's activity

What speed do you get at Q8 on AMD Ryzen™ AI Max+ 395

Where is vision support like previous devstral version?

New Chat Template + Tool Calling Fixes as of 05 Aug, 2025

Will there be fix new proper model or jinja template fix for Roocode/Cline/Goose etc.?

MMProj Validation Error (MMProj file baseName "Mistral-Small-3.1-24B-Instruct-2503" does not match model name "Devstral-Small-2507".)

Well I tried the new DevQuasar Kwaipilot.KAT-Dev.Q8_0.gguf - Lmstudio and Jan.ai errors out:

This one's great!

the instruct following is poor

How much VRAM is required for this one

KAT-DEV 33B no longer working with Jan.ai 0.7.2 - Cline / LMstudio v0.3.31b4 - Cline

Is KAT-Dev supposed to be that much slower than QWEN?

KAT-DEV Q4 - Jan.ai - Roocode / Cline does not work Lmstudio crashes too (SOLVED)

coder vs yoyo-v4 vs thinking vs instruct in compex programming task

Request GGUF: Kwaipilot/KAT-Dev (62.4% (!) on SWE-Bench Verified having just 32B)

There is new KAT-Dev-72B-Exp

This model gives errors in cline, but new official one does not

why Kat-Dev worse then Qwen3-coder 30B?

errors out with roocode