@raymundoojeda1: Qwen3.8 27B on Ninfer is insane! 244 Tok/s #ai #localai #nvidia #fyp

Raymundo Ojeda
Raymundo Ojeda
Open In TikTok:
Region: US
Thursday 17 September 2026 21:20:21 GMT
9456
263
16
27

Music

Download

Comments

ryuk.246
Ryuk yêu em :
my 3080 12vram 36gb ram can go 900tok/s on qwen 3.8 30b. but only 9k context window until everything breaks 😭
2026-09-18 13:43:37
0
quandinglus
Quandinglus Lemar Tickleton :
this is why the 5090 is 9000 dollars?
2026-09-18 10:55:15
6
benot479
Benoît :
Same here
2026-09-18 11:48:58
0
neutrinotau
XYZ :
What CTX did u use? I get 96 tok/sec with a 128k ctx
2026-09-18 04:41:24
1
weirdturnedpr0
weirdturnedpr0 :
what chat ui are you using?
2026-09-17 21:49:24
0
scarlet_alibi
Alibi :
250-270k is still 160-180 tps on ninfer 5090
2026-09-18 10:08:59
0
onepunmang
DaDawwg •ו :
chatgpt told me it can do like 30 tokens/s xD brooo whats your setup?
2026-09-18 11:17:57
0
escob.art
escob.art :
It’s great. Managing my openclaw on a m1 mac studio from a 5090 over network 🦞😻
2026-09-18 01:05:39
0
sreckojovancevic
Srecko :
context widrh?
2026-09-18 05:34:26
0
applefarmer97
Isabella Farmer :
The local area are gonna go away. I hate to tell you all you have to do is do a deep dive with Gemini and tell you.
2026-09-17 22:43:25
0
brzhifi
ruslanmulyk :
Looks strong my friend😃
2026-09-17 22:13:25
0
lazercube
Lazercube :
4090 also available, it's a fork on git: ninfer-4090
2026-09-18 06:32:57
0
To see more videos from user @raymundoojeda1, please go to the Tikwm homepage.

Other Videos


About