GoKawiil
Show HN: Lumabri – Run Moe Models on a P2P Swarm with Colibri
(news.ycombinator.com)
1.
2.
Rotary GPU: Exploring Local Execution for Large MoE Models Under Limited VRAM
(news.ycombinator.com)
3.
Real-time LLM Inference on Standard GPUs: 3k tokens/s per request
(news.ycombinator.com)