Guide
How much does AI model usage cost for an agent?
Model usage is billed by tokens: the text the model reads in and writes out. There is no flat monthly fee; the cost tracks how much the assistant works. A light helper can cost a dollar or two a month, while an agent working all day can cost many times more, even on the same model.
GPT-OSS 20B across usage levels
| Usage | Model cost / mo |
|---|---|
| A few questions and small tasks each day. | $0.30/mo |
| Daily use, plus some scheduled tasks. | $1.00/mo |
| Working for you most of the day, including web tasks. | $3.00/mo |
Token cost only, at current verified prices. A premium model can cost many times more for the same work. See the full model list. Add a host for the full-stack total.
Why usage, not a subscription
Hosting keeps the assistant online for a fixed monthly fee. Model usage is separate and variable: every reply the model reads and writes costs tokens. If the assistant works often, model usage can quietly become the largest line on the bill, which is why AIStackPicker estimates it instead of hiding it.
What changes the cost
- Compute: where it runs: a cloud host, a rented GPU, or hardware you own.
- Model usage: how much it reads and writes, and how often it runs.
- Runner: the software that serves the model, or runs the agent loop and its tools.
- Capabilities: browser automation, web search, memory, or scheduled work.
- Required extras: security, delivery, monitoring, or managed-host fees.
- Freshness: whether each price is verified, aging, stale, or disputed.
How AIStackPicker checks it
The figures above use current verified token prices at three usage levels, not a vendor headline rate. The Builder applies the same usage model to whichever model you pick, so you can compare like for like.
Related questions
Related guides
Every price is sourced and dated at its row.