Run full Kimi K3 on a single device. And an OpenAI-compatible API server for local chat and coding agents.
Terminal score: 0–1000 raw, weighted across 4 dimensions. Public score: 0–10 normalized (shown in the 30-day stars chart above).