# Consumption and Rate Limits
## Overview
The cost of using the API is based on token consumption, with different prices based on token category:
- Prompt text tokens
- Audio tokens
- Image tokens
- Cached prompt tokens
- Completion tokens
- Reasoning tokens
## Rate Limits
For each tier, there is a maximum amount of:
- **Requests per minute (RPM)**
- **Tokens per minute (TPM)**
These limits ensure fair usage by all users of the system.
### Rate Limit Error
Once your request frequency has reached the rate limit, you will receive error code **429** in response.
```json
{
"error": {
"message": "Rate limit exceeded",
"type": "rate_limit_error",
"code": 429
}
}
```
### Requesting Higher Limits
To request a higher rate limit, email **support@x.ai** with your anticipated volume.
## Token Pricing
Requests using xAI-provided server-side tools are priced based on:
- Token usage
- Server-side tool invocations
Costs scale with query complexity.
### Cached Tokens
Using cached input tokens can significantly reduce costs. Tokens that match previous requests may be served from cache at a reduced rate.
## Pricing by Token Type
| Token Type | Description |
|------------|-------------|
| Prompt tokens | Input text tokens |
| Completion tokens | Generated output tokens |
| Reasoning tokens | Tokens used for model reasoning |
| Image tokens | Tokens for image processing |
| Audio tokens | Tokens for audio processing |
| Cached tokens | Previously processed tokens (discounted) |
## Model-Specific Pricing
Different models have different pricing tiers. Check the [Models and Pricing](https://docs.x.ai/docs/models) page for current rates.
## Monitoring Usage
### Usage Explorer
You can see usage with grouping to easily compare consumption across groups.
For example, compare consumptions across test and production API keys by selecting **Group by: API Key**.
### Usage Dashboard Features
- Real-time usage tracking
- Historical usage data
- Cost breakdown by model
- API key-level analytics
## Billing
### Prepaid Credits
- Pre-purchase credits for your team
- API consumption deducted from remaining prepaid credits
- No overage charges until credits exhausted
### Monthly Invoiced Billing
- xAI generates monthly invoice based on API consumption
- Used when prepaid credits are not available
- Billed at the end of each billing period
## Best Practices for Cost Optimization
### 1. Use Appropriate Models
Choose the right model for your task:
- Simple tasks → grok-4-1-fast-non-reasoning (cheaper, faster)
- Complex tasks → grok-4 (more capable)
### 2. Optimize Prompts
- Be concise in system prompts
- Avoid redundant context
- Use efficient prompt structures
### 3. Leverage Caching
- Repeated prompts benefit from token caching
- Structure requests to maximize cache hits
### 4. Monitor Usage
- Regularly check the Usage Explorer
- Set up alerts for usage thresholds
- Track costs by API key/project
### 5. Use Streaming Wisely
- Streaming doesn't change token costs
- But can improve perceived latency
## Handling Rate Limits
### Exponential Backoff
```python
import time
import random
def make_request_with_backoff(client, **kwargs):
max_retries = 5
base_delay = 1
for attempt in range(max_retries):
try:
return client.chat.completions.create(**kwargs)
except Exception as e:
if "429" in str(e) and attempt < max_retries - 1:
delay = base_delay * (2 ** attempt) + random.uniform(0, 1)
print(f"Rate limited. Waiting {delay:.2f}s...")
time.sleep(delay)
else:
raise
```
### Request Queuing
```python
import asyncio
from collections import deque
class RateLimitedClient:
def __init__(self, client, rpm_limit=60):
self.client = client
self.rpm_limit = rpm_limit
self.request_times = deque()
async def make_request(self, **kwargs):
# Wait if at rate limit
while len(self.request_times) >= self.rpm_limit:
oldest = self.request_times[0]
if time.time() - oldest > 60:
self.request_times.popleft()
else:
await asyncio.sleep(1)
self.request_times.append(time.time())
return self.client.chat.completions.create(**kwargs)
```
## Status Monitoring
If there is an ongoing service disruption:
- Visit https://status.x.ai for latest updates
- RSS feed available at https://status.x.ai/feed.xml
