What does an AI server cost to build?
The accelerator sets most of the bill. Learn how VRAM, memory bandwidth, power headroom, cooling and your own tariff decide what the rest costs.
#self-hosted-llm
Filtering by topic #self-hosted-llm · clear
Zanus AI vs building your own private AI server
Zanus AI sells a no-code AI appliance priced by quote. What the same private stack takes to build on a rented GPU server, and when an API is the better buy.
Claude plans when your payment is declined
Every Claude plan price in USD as of September 2026, why a card gets declined in an unsupported country, and the open model option you can still run.
Run Gemma 4 on a VPS: sizing and speed
Which Gemma 4 tag fits your VPS RAM, what each one costs in tokens per second on CPU, and the sizing arithmetic to check it before you pull 20GB.
The Ollama API and port 11434 explained
Curl your Ollama server on port 11434, read what a refused connection means, tour the API endpoints, and set OLLAMA_HOST without exposing the box.
Ollama concurrency: NUM_PARALLEL and MAX_QUEUE
The second request waits, or it is refused: how OLLAMA_NUM_PARALLEL and OLLAMA_MAX_QUEUE decide which, and why every parallel slot costs you VRAM.
Run GLM 5.2 on a VPS with Ollama
GLM 5.2 is cloud only in Ollama's library. Here is the GLM model that actually fits a VPS, and the RAM each quantisation needs on your own box.
Run the llama.cpp server on a VPS
Build llama-server from a pinned tag, serve GGUF models on the OpenAI-compatible API, bind it to localhost, and run it under systemd with memory limits.
Run DeepSeek V4 Flash on a VPS
DeepSeek V4 Flash on Ollama is cloud only. Get the real disk and RAM numbers for every GGUF build, and see which VPS plan can actually load it.
What it takes to self-host Kimi K3
Kimi K3 is 2.8 trillion parameters. Here is the VRAM arithmetic, the KV cache math, and the three honest ways to run it without a 32 GPU cluster.