What are the Grok 4.7 rate limits on the xAI API?

On the xAI API, grok-4.7 runs from 150 requests a second and 50M tokens a minute at Tier 0 to 500 requests a second and 100M tokens a minute at Tier 4, which takes $5,000 of spend.

Checked against the lab's own pages on September 26, 2026. Limits change often, so the date matters.

On the xAI API, grok-4.7 is limited two ways at once: requests per second and tokens per minute. Both climb with your team's tier, and SpaceXAI prints the whole table on its rate limits page. A new team starts at Tier 0 with 150 requests a second and 50 million tokens a minute, which is also what the model's own page shows.

Tier 0 ($0): 150 RPS, 50M TPM
Tier 1 ($50): 172 RPS, 53M TPM
Tier 2 ($250): 208 RPS, 60M TPM
Tier 3 ($1,000): 312 RPS, 74M TPM
Tier 4 ($5,000): 500 RPS, 100M TPM

How the tiers work

Your tier comes from what your team has spent on the xAI API since January 1, 2026, counting prepaid credit purchases and paid invoices. Tiers unlock on their own as spend goes up, and they never go back down. Past Tier 4 there's Enterprise, available on request.

Grok 4.7 doesn't get its own row of numbers. The table gives grok-4.5, grok-4.6 and grok-4.7 identical limits at every tier, so moving up from an older model doesn't change your headroom. Older models like grok-4.3 start far lower, at 37 requests a second.

What counts, and when it resets

The per-second limit is your per-minute request budget divided by 60, so you can't burn a minute's worth of requests in one second. Tokens are counted per minute and include everything the request uses:

prompt tokens, including images and audio
completion tokens
reasoning tokens
cached prompt tokens, which cost less but still count toward TPM

That last one catches people out. Caching lowers the bill without lowering the rate you're metered at.

What happens when you hit it

Exceeding any limit returns a 429 Too Many Requests error.

From Rate Limits | SpaceXAI Docs

SpaceXAI's advice is exponential backoff: wait, double the wait on each retry, and give up after a handful. Its sample code tries up to five times before it stops.

There are three ways up. Spend more and the tier moves by itself. Ask through the xAI Console for higher limits without extra spend or beyond Tier 4. Or email sales for enterprise capacity. You can see your team's own tier and per-model limits on the Models page in the Console.

These are API limits. The page doesn't cover Grok 4.7 inside the Grok app or in tools like Cursor, which set their own quotas.