"Traditional approaches to rate limiting APIs won’t work effectively for AI agent consumers, so some API providers have shifted to adaptive rate limiting (ARL). For example, DeepSeek employs a more dynamic and adaptive approach to rate limiting its API compared to other LLM API providers currently.
The concept of adaptive rate limiting isn’t new, but it’s evolving to address new API usage scenarios that include AI agents. Modern ARL involves a set of principles, tools, and techniques that allow systems to adjust rate limits dynamically based on context and real-time insights. It includes a combination of approaches:"
https://nordicapis.com/how-ai-agents-are-changing-api-rate-limit-approaches/