Guides

Guides for running an LLM or AI agent

Start with the question you actually have. Each guide is short, sourced, and keeps the cost models separate (API tokens, rented GPU time, and local hardware), then links into the Builder when the decision gets specific.

How much does it cost to run an LLM?

The three cost models kept separate: API tokens, rented GPU time, and local hardware.

How much does it cost to run an AI agent?

Why agents cost more than chat: steps, tool calls, retries, memory, and always-on hosting.

Local LLM vs cloud API: what actually changes the cost?

What a cloud API bills for, what local shifts onto hardware and setup, and when each one wins.

Can my computer run a local LLM?

Whether your machine can run a local model is mostly a memory question: RAM, VRAM, unified memory, quantization, and context.

GPU hosting vs model API: which is cheaper?

Why GPU cost depends on hours and utilization, while API cost depends on tokens.

Why isn't local AI always free?

Local inference drops the token bill, but hardware, power, setup, heat, and reliability still cost.

What is a model runner?

The software that loads a model and serves inference, not the model, the interface, or the agent runner.

What is an agent runner?

The software that runs the agent loop (steps, tool calls, retries, memory) versus a model runner that only serves the model.

LM Studio vs Ollama: which should you use?

Two Model runners for local models (a desktop app vs a command-line service) and which fits which need.

OpenRouter vs direct provider APIs: what's the difference?

Both reach a hosted model over an API and bill by tokens. One is an aggregator with one key for many providers, the other is the model maker's own account.

What is the cheapest always-on AI agent setup?

Why a small always-on host plus a light API model is usually the low-cost path, and when a rented GPU is the wrong tool.

What is the best interface for an AI agent?

Why the interface is separate from the runners, and how the right one depends on who uses it and whether it must stay online.

What is the cheapest always-on AI assistant setup?

The lowest verified full-stack cost to stay always on, itemized and dated.

How much does a self-hosted AI assistant cost?

Fixed host cost, model usage, setup effort, and privacy tradeoffs, separated.

What does it cost to host an AI agent?

Why hosting is only one line item, and what the model adds on top.

Should your AI assistant run locally or in the cloud?

When your own computer is enough, and when you need a VPS or managed host.

What does an AI agent need to work the web?

How web browsing affects memory, host, security, and cost.

How much does AI model usage cost for an agent?

Token cost at light, moderate, and heavy daily usage, with real numbers.

What is the best VPS for an AI assistant?

The cheapest rented servers that meet the memory floor, compared and dated.

What is the best no-setup AI assistant host?

Managed hosts that run the machine for you, and what the convenience costs.

Build your setup Browse all stack parts

Loading the latest dated prices…