@tuaniiht: AirLLM enables LLaMA 3.3 70B on MacBook or gaming PC by layer-by-layer loading and FlashAttention, delivering near-flat memory for an open-source on-device AI revolution #AirLLM #LocalLLM #LLaMA70B #OpenSourceAI #EdgeComputing #fyp #foryou
The GitHub repository hadn’t been updated in 3 years.
2026-06-04 13:44:36
4
Roy Anderson :
instead of TPS you get MPT.
2026-07-24 02:33:36
0
sicksevenfive :
Flash attention isnt nee btw
2026-07-01 07:26:48
0
Athion :
Hint : you can’t. This thing is a research experiment and nothing more. At best, its output is measured in tokens per MINUTE. It’s completely unusable for anything other than saying “look, I got a 70b to run on a 4gb card”.
2026-06-28 20:55:16
1
Globaluser :
It delay the response by 45s
2026-08-31 12:15:20
0
Anony Mous :
ollama?
2026-06-23 00:57:54
0
thestonedgamer420 :
it's slow as balls tho
2026-06-03 13:35:05
1
iCreatr :
Nonsense
2026-06-19 20:06:05
0
To see more videos from user @tuaniiht, please go to the Tikwm
homepage.