Two NVIDIA GeForce RTX 3090 GPUs and Intel Optane run 2.8-trillion-parameter Kimi K3 at 3.11 tokens/sec
TweakTown | 21.11.2026 21:35
A PC enthusiast has managed to run Moonshot AI's massive 2.8-trillion-parameter Kimi K3 language model locally at 3.11 tokens per second, using two NVIDIA GeForce RTX 3090 GPUs, an Intel Xeon server, and Intel Optane persistent memory.