Term
TPM/RPM
TPM (tokens per minute) and RPM (requests per minute) are the two common units in which AI API providers express their rate limits — TPM caps the token volume, RPM caps the number of requests per minute.
TPM/RPM — explained in detail
TPM and RPM are the two units of measure that AI API providers use to define their usage limits. RPM stands for “requests per minute” and caps how many individual requests an account may send per minute. TPM stands for “tokens per minute” and caps the total of processed tokens per minute — usually input and output tokens combined.
Both limits generally apply at the same time: if an application hits the RPM limit first, that one blocks; if it consumes very many tokens per request, the TPM limit kicks in first. Which of the two axes throttles first depends on the usage pattern. Many small requests tend to hit the RPM limit; a few very large prompts tend to hit the TPM limit. Some providers add longer windows such as tokens per day (TPD).
When either limit is exceeded, the API typically responds with HTTP status code 429 and often a Retry-After hint. With most providers the TPM and RPM values rise with usage and payment history (a tier system).
Example / Practical use
A service has a limit of 500 RPM and 90,000 TPM. If it sends 500 short requests of 100 tokens each, it stays under the TPM limit on token volume (50,000) but hits the RPM limit exactly. If it instead sends only 50 requests of 3,000 tokens each, the RPM limit is no issue, but 150,000 tokens exceed the TPM limit. Anyone planning for this picks smaller prompts for simple bulk tasks and batches work through the batch API, which uses a separate quota.
Distinction from related terms
TPM and RPM are the units; the overarching concept is the rate limit. Neither should be confused with the context window, which limits the token count within a single request — TPM, by contrast, sums across many requests per minute. TPM is also not a cost metric: what processing costs is expressed by the price per million tokens (see cost per MTok).
Entdecke mehr
Headless without an API bill — how do you reach the best AI models for automation in 2026?
Provider comparison mid-2026: who has a headless mode, whose subscription still covers it — and why BYOK is the most stable foundation.
GlossarEU AI Act (KI-Verordnung)
Regulation (EU) 2024/1689 is the EU's first comprehensive law on artificial intelligence. It regulates AI systems by risk in four tiers, from prohibited practices to minimal risk, and sets obligations for providers and deployers.
LexikonFunction Calling / Tool Use
How an LLM uses tools: define a tool as a schema, the model picks the function and arguments, the result returns to the chat — the basis of every agent.