Kimi K3, Moonshot AI's 2.8 trillion parameter open-weights MoE model, requires roughly 1.5 TB of RAM to run inference — about 1.4 TB for weights alone at native MXFP4 4-bit precision. Running it demands at minimum 8x next-gen GPUs (NVIDIA B300 or AMD MI350X) capable of native FP4 execution, since older Hopper/MI300X hardware dequantizes at runtime and loses the speed benefits. The cost math shows self-hosting on rented GPUs runs ~$38/hr (~$27,800/month), making serverless inference far cheaper for most teams unless they sustain 40+ heavy concurrent users generating 50M+ tokens/month. Key architectural features like Kimi Delta Attention (KDA), Stable LatentMoE, and Attention Residuals improve scaling efficiency ~2.5x over K2. A custom vLLM build with K3-specific code is required for self-hosting. DigitalOcean offers both serverless inference at $3/$15 per 1M tokens and managed GPU options with MI350X and B300 hardware.
