API Rate Limits
Ensure reliable integration and platform stability by understanding how NexusAI manages request throttling, burst allowances, and tier-based quotas.
📊 How Rate Limiting Works
NexusAI uses a sliding window token bucket algorithm to distribute fair usage across all clients. Requests are evaluated against both per-minute and daily thresholds. When a limit is exceeded, the API returns a 429 Too Many Requests status with retry instructions.
Standard Limits by Tier
| Tier | Requests / Minute | Requests / Day | Concurrent Connections | Burst Allowance |
|---|---|---|---|---|
| Free | 60 | 5,000 | 2 | 10 req / 5s |
| Professional | 600 | 150,000 | 10 | 50 req / 5s |
| Enterprise | Custom | Unlimited* | Unlimited | Custom |
*Enterprise daily quotas are negotiated during onboarding and tied to your dedicated infrastructure allocation.
🔑 Response Headers & Retry Logic
Every NexusAI API response includes rate-limit metadata. Monitor these headers to gracefully handle throttling without breaking your application.
// Sample Response Headers
X-RateLimit-Limit: 600
X-RateLimit-Remaining: 142
X-RateLimit-Reset: 1709283600 (Unix timestamp)
Retry-After: 23 (seconds, only on 429)Recommended Retry Strategy
- Exponential Backoff: Start with a 1s delay, doubling up to a max of 30s on consecutive 429s.
- Jitter: Add random jitter (±10-20%) to prevent thundering herd effects.
- Circuit Breaker: If failure rate exceeds 40% in a 60s window, pause requests for 30s.
- Prioritize Reads: Cache responses where possible to reduce write-heavy API calls.
429 responses or Retry-After headers may trigger temporary API key suspension for up to 24 hours to protect platform stability.⚡ Endpoint-Specific Throttling
Certain heavy-compute endpoints enforce stricter limits to balance shared GPU/TPU resources:
| Endpoint | Default Limit | Notes |
|---|---|---|
POST /v1/chat/completions |
100 req/min | Token-heavy; counts toward tier base limit |
POST /v1/embeddings |
300 req/min | Batched payloads reduce overhead |
POST /v1/audio/transcriptions |
50 req/min | File upload + processing pipeline |
POST /v1/fine-tunes |
5 concurrent jobs | Queue-based; not rate-limited by time |
Use the X-Endpoint-Limit header in responses to check per-route availability.
📈 Requesting Higher Limits
Need to scale beyond your current tier? NexusAI offers flexible quota expansions without requiring immediate plan upgrades.
- Temporary Burst: Apply for 24-72 hour quota increases via the dashboard during events or launches.
- Dedicated Throughput: Enterprise clients can provision isolated API gateways with guaranteed TPS.
- Webhook Offloading: Switch to async processing with webhooks to bypass synchronous rate limits.
Ready to scale your API usage?
Contact our solutions team to design a custom rate limit profile that matches your traffic patterns.
Request Quota Increase →