You do not need a paid plan to learn Grok. Start on the free allowance, do one real piece of work, and only then decide whether a subscription or the API makes sense. This lesson gets you to that first result in about fifteen minutes.
First 15 minutes with Grok
- 1
Open Grok
In the X app or website, look for the Grok entry in the side navigation. You can also use the standalone Grok web app or mobile app, which sign in with the same account. Either route uses your free allowance.
- 2
Check which model you are on
The model picker sits near the input box. On the free allowance you may be limited to a lighter model. Note the name; when you compare answers later you want to know what produced them.
- 3
Decide on real-time X context
Grok can search X posts when answering. This is great for current events and terrible for careful research. Turn it off, or tell Grok not to use it, when you want a plain answer based on its training.
- 4
Review the privacy setting
In your X settings there is a control for whether your posts and interactions are used to train Grok. Decide deliberately. This does not change what you pay, but it changes what you give away.
- 5
Do a first real task
Paste a long thread, article, or email and ask for a five-bullet summary with one recommended action. Compare the result with what you would have written. That is the quality bar you are paying for.
- 6
Note when you hit the limit
If the free allowance runs out during normal use, write down how long you had been working. That number is your first data point on whether a paid plan is worth it.
Interface tour
The chat window itself is familiar if you have used any assistant. Three parts matter more than the rest. The model picker decides quality and how fast your allowance burns. The conversation history is your context: everything earlier in the thread is re-read on each turn, so long threads get slower and, on the API, more expensive. The attachment and search toggles decide whether Grok reads a file or X posts before answering.
Settings that matter
- Model choice. Default to the lighter option. Switch up only when an answer is not good enough.
- X search. On for news and trends, off for analysis and drafting.
- Training opt-out. Set once in X privacy settings.
- New chat per task. Old context confuses the model and, on the API, costs money every turn.
If you are here as a developer, skip the app after your first task and go to Lesson 7. The API gives you model choice per request and a bill that matches what you use, which is usually the cheaper route for scripts and automation.
Knowledge check
Why does starting a new chat for each new topic reduce cost on the API?