• Llama 3.1 405B now runs at 969 tokens/s on Cerebras Inference https://cerebras.ai/blog/llama-405b-inference 156 comments 19/11/2024