@dr_cintas: You can now run a 70 billion parameter AI model on a graphics card with just 4GB of VRAM

dr_cintas
dr_cintas
Open In TikTok:
Region: US
Wednesday 22 July 2026 14:15:43 GMT
4669
197
5
30

Music

Download

Comments

anonymous_student_1
OmgWhoCaresWhatMyUsernameIs :
Stop with the AirLLM propaganda, it's useless. It runs at very very slow speeds (1 token per second) it's the same as spilling your LLM into "Swap Ram" it can run but it's not gonna be usable.
2026-07-22 15:12:29
2
aoneesh.sharma
Aoneesh :
Can AirLLM and Colibri be run together?
2026-07-23 10:03:55
0
aifastlaners
aifastlaners :
1 token/second?😭😭😭
2026-07-22 14:46:17
1
To see more videos from user @dr_cintas, please go to the Tikwm homepage.

Other Videos


About