@analogalok@selogerkkk When you say without any kv cache quantization do you mean running f16 kv cache over q8 kv cache? I'd be interested in seeing your config / params for running this model
@rossdrummer@AlexFinn Calm the fuck down. Anthropic will re release it when they have a system that can verify if you are a US citizen using the model. Since they currently have no system in place for ID verification they had to remove model.