@momus.ai: QWEN 3.8 next flash on a DGX spark at 524k context. Two QWEN3.8 27b subagents at 256k and 64k context. Deepseek harness with custom plugins #qwen #deepseek #localai #nvidia #amd

Momus - Local AI Lab
Momus - Local AI Lab
Open In TikTok:
Region: US
Tuesday 01 September 2026 07:09:10 GMT
95076
2637
299
216

Music

Download

Comments

yikator1
fuyu愛情 :
why does yall need local ai
2026-09-12 04:30:36
2
mysterio0922
Mysterio :
lol my 5070 runs this at Q4_K
2026-09-02 00:10:15
30
aryanalkhair
aryan :
hey i have 4gb of vram and 16 gigs of ram, what can i run?
2026-09-02 04:24:47
14
streetshortstories
streetshortstories :
4× RTX 3090 — all watercooled. Any Idea? 😏🔥
2026-09-02 20:36:53
19
brandonbarcus6
Brandon Barcus :
1100w? I thought my vintage cards running qwen3.8 was bad. Only get 44t/sec
2026-09-01 23:13:01
17
cesicoma
cesicoma :
Its my fuxking dream setup with deepseek harness
2026-09-04 18:07:45
7
omg.hilarious6
omg hilarious 😂 :
im looking to spend $5-6k
2026-09-01 17:37:07
7
adam.stevens170
Adam Stevens :
Can I pay you to use ur ai?
2026-09-25 01:46:55
0
sergio1492_
Sergio :
How does the dgx compare to the mac mini and whatever gpu you have for the 27b model in prefill and decode tps? Curious and wondering if i should buy a dgx or mac studio…
2026-09-03 15:32:47
1
mrkcweld6061
JayFabWerks :
Ah, this is what Qwen is supposed to look like when not being ran on my laptop with a tired 3050 and 4GB VRAM😂😂😂
2026-09-01 10:52:43
19
alib_dz
alib_dz :
@alib_dz: Bro i have rx 6700xt 12 vram and 16gb ram how much more i need to run qwen 3,8 on ny pc please
2026-09-01 23:43:00
0
_chubbz_8
_Chubbz_ :
Why are you running two 27B’s instead of running one with more parallels?
2026-09-03 09:12:46
7
aarronphotography
Aarron :
I have 12GB of VRAM and 128GB 3600Mhz DDR4, how far can I get with that setup
2026-09-10 05:33:58
1
con.freq
Conspiracies & Frequencies :
Too much money, not enough knowledge.
2026-09-03 12:17:07
17
anchorboo
jey jey jeeey :
how's the pp and tg speeds?
2026-09-04 22:05:57
1
ionegamer
IONE Gamer :
эта модель на спарке хорошо тянет сложные агентские задачи?
2026-09-04 06:16:34
0
weweresafehere
weweresafehere :
И все для генерации аниме девочек
2026-09-02 20:05:53
5
skalopendr
Skalopendr :
сейчас пишу приложение для смартфона, для работы с qwen 2.5 1.5b локально. смартфон с 4+ Гб оперативной памяти😉
2026-09-02 10:54:10
3
mocha_kafe
mocha_kafe :
I got 8 vram and 32 ram. What can I do?
2026-09-02 06:33:03
1
dkw999_
dkw999_ :
so its cheaper to use claude max??
2026-09-01 21:34:14
2
ajsdumb
Reese<3 :
I got a 9800x3d 128gb of ddr5 ram at 6000mhz cl36, rx 9070xt 16gb vram, 4tb of ssd what can i do to make money with my pc? Genuine question
2026-09-24 18:40:00
0
zeidexsoft
ZeidexSoft :
For example, if you want to run local models with the possibility of using them in a local network by 100 users, is the budget of 20 to 30 thousand dollars enough to buy devices, of course, without network devices
2026-09-17 02:08:42
0
user92183144626195
user92183144626195 :
For full stack developer, what is better? A Spark 128gb or Mac Studio M5 ultra 92 gb ?
2026-09-22 17:13:48
0
rwear39
ACE🍺 :
I found cheap server equipment used 250 gb ram thread ripper on it. my Guess next few models optimize more to smaller PC maybe just maybe it be worth it . 1500 PC + old parts.
2026-09-03 18:02:32
1
To see more videos from user @momus.ai, please go to the Tikwm homepage.

Other Videos


About