[CamperBob2]

It's an MoE model, so it might not be that bad. The deployment guide at https://huggingface.co/moonshotai/Kimi-K2-Thinking/blob/main... suggests that the full, unquantized model can be run at ~46 tps on a dual-CPU machine with 8× NVIDIA L20 boards.

Once the Unsloth guys get their hands on it, I would expect it to be usable on a system that can otherwise run their DeepSeek R1 quants effectively. You could keep an eye on https://old.reddit.com/r/LocalLlama for user reports.

[wg0]

Are such machines available in the A class clouds such as Azure/AWS/Google?