LLM API Rate Limits and Pricing Tiers: A Complete Guide
Published 2026-08-02 · Updated 2026-08-02
LLM API pricing is not just about per-token costs. Rate limits (requests per minute, tokens per minute) and tier-based pricing systems determine how much you can actually use — and how much you pay at scale. This guide explains how these systems work and how to plan around them.
What are rate limits?
Rate limits restrict how many requests (RPM) or tokens (TPM) you can send within a time window. They exist to ensure fair usage and prevent abuse. When you hit a rate limit, the API returns a 429 error and you must wait before retrying.
Rate limits are separate from pricing tiers. A higher tier usually means higher rate limits, but the per-token price may also change.
OpenAI tier system
OpenAI uses a tiered system (Tier 1 through Tier 5) based on your total spend. Higher tiers unlock higher rate limits and access to more models.
| Tier | Qualification | GPT-4o RPM | GPT-4o TPM | GPT-4o mini RPM |
|---|---|---|---|---|
| Tier 1 | $5+ spend | 500 | 800k | 500 |
| Tier 2 | $50+ spend | 2,500 | 2M | 2,500 |
| Tier 3 | $100+ spend | 5,000 | 4M | 5,000 |
| Tier 4 | $250+ spend | 10,000 | 8M | 10,000 |
| Tier 5 | $1,000+ spend | 30,000 | 30M | 30,000 |
Rates are approximate and vary by model. Check OpenAI's rate limits documentation for current values.
Anthropic tier system
Anthropic uses a simpler tier system based on your billing tier. Rate limits scale with your usage level and can be increased by contacting sales.
| Tier | Claude Sonnet 4 RPM | Claude Sonnet 4 TPM | Claude Haiku 4.5 RPM |
|---|---|---|---|
| Free | ~50 | ~40k | ~50 |
| Build | ~1,000 | ~800k | ~1,000 |
| Scale | ~5,000 | ~4M | ~5,000 |
Anthropic's rate limits are not publicly documented in detail. Values above are approximate based on community reports. Contact Anthropic sales for exact limits.
Google Vertex AI quotas
Google Cloud uses project-level quotas for Vertex AI and Gemini API. Quotas are measured in requests per minute (RPM) and tokens per minute (TPM) per project, not per API key.
| Model | Default RPM | Default TPM | Increase possible? |
|---|---|---|---|
| Gemini 2.5 Flash | 2,000 | 4M | Yes, via quota increase request |
| Gemini 2.5 Pro | 1,000 | 2M | Yes, via quota increase request |
| Gemini 3.1 Flash Lite | 3,000 | 6M | Yes, via quota increase request |
Google quotas are per-project and can be increased by filing a quota increase request in the Google Cloud Console. This is free but may take 1-2 business days to approve.
How rate limits affect cost planning
Rate limits affect your cost in two ways:
- Throughput ceiling: If your workload exceeds the rate limit, you must either spread requests over time, use multiple API keys/projects, or upgrade to a higher tier.
- Tier-based pricing: Some providers (notably OpenAI) change per-token pricing at higher tiers. Tier 4+ models may have different rates than Tier 1.
Use the cost calculator to estimate your monthly volume, then check if your target provider's rate limits can handle it.
Strategies for working within rate limits
| Strategy | When to use | Tradeoff |
|---|---|---|
| Batch API | Offline workloads that can wait 24h | 50% cost savings, but no real-time response |
| Multiple API keys | Distributed workloads across teams | Each key has its own rate limit pool |
| Request queuing | Bursty workloads with predictable peaks | Adds latency, requires retry logic |
| Model cascading | When primary model is rate-limited | Fallback to cheaper/faster model |
| Contact sales | Enterprise workloads needing high RPM | May require minimum spend commitment |
Pricing data sourced from our catalog. Check data sources for provenance and freshness.
Start here
Use the cost calculator to estimate your monthly volume, then check your target provider's rate limit documentation. If you're hitting limits, consider batch processing or model routing to stay within quotas.