benchrInput / 1M$2base
Output / 1M$6base
Cached / 1M$0.50base
Context500Kno numeric output cap
Official price table
| Tier | Input / 1M | Output / 1M | Cached input / 1M |
|---|---|---|---|
| Prompt below 200K | $2 | $6 | $0.50 |
| Prompt at or above 200K | $4 | $12 | $1.00 |
The threshold reprices the whole request
xAI says long-context rates apply to every token in a request once its prompt reaches 200K tokens. The separately announced fast variant is a different option and is not what the second row represents. Set maximum turns, time, output, and tool budgets for long-running agents.
Frequently asked questions
What does Grok 4.6 cost below 200K?
$2 input, $6 output, and $0.50 cached input per million tokens.
What happens at 200K prompt tokens?
The whole request uses the $4 input, $12 output, and $1 cached-input long-context rates.
What is the maximum output?
The checked guide says there is no text-output limit but does not give a numeric token cap; benchr does not invent one.