#LLM inference cost

2 articles tagged

Illustration weighing self-hosting an AI model against calling a cloud API

AI & Cloud Infrastructure · Aug 1, 2026 · 12 min read

Kimi K3 self-hosting vs API in 2026: 1.68 TB of VRAM and the break-even math

Kimi K3's open weights are 594 GB in BF16 and need about 1.68 TB of VRAM to serve; the API lists at $3/$15 per 1M tokens. Here is the 2026 self-host vs API break-even math with real GPU rates.

Manu Shukla

#Kimi K3 #self-hosting #LLM inference cost

Glowing GPU server module on a dark studio surface, representing DeepSeek V4 self-hosting cost

Cloud & FinOps · Jul 28, 2026 · 14 min read

DeepSeek V4 self-hosted vs API in 2026: the real GPU cost and break-even math

DeepSeek V4 is both a cheap API and an open-weight model you can self-host. Here is the 2026 cost math: Flash at $0.14/$0.28 per 1M tokens, H100/H200 rental rates, and the GPU utilization point where self-hosting pays.

Manu Shukla

#DeepSeek V4 #LLM inference cost #self-hosting