B blog.kog.ai Real-time LLM Inference on Standard GPUs: 3k tokens/s per request 219 points NicoConstant 5 months ago 97 comments
