We are really wondering how many request we can do at 1 minutes. What are exactly request rate limits for all endpoints? We are asking that because we always get some rate limit errors 429. However we request (approximately 20) affordable counts in our opinion. Thanks for any help.

It is 50req/min

Hey! The 429s are likely tied to your usage tier, not just request count. Even 20 req/min can hit limits if you're on Tier 1, or if you're exceeding token-per-minute (TPM) caps. Quick wins: check your tier in the Anthropic console and add exponential backoff on 429s. Happy to dig into your setup and fix this properly let me know! @muratsk

Hey Adam thanks for your quick response, it helps me a lot (i had no idea about TPM and Anthropic, :)) but still there are some issues in my mind. We are currently building a real-time voice assistant using your platform and have been encountering frequent 429 rate limit errors during active conversations as you know. We initially assumed this was related to request-per-minute limits, but based on your feedback and our observations, it appears that we are more likely hitting token-per-minute (TPM) constraints, especially given the nature of voice interactions. In our use case, each conversation involves multiple back-and-forth turns, and each turn triggers an LLM call with accumulated conversation context. As a result, even a small number of concurrent calls can generate significant token throughput. We are observing the following issues: • Intermittent 429 errors even at relatively low request volumes (~20 req/min) • Noticeable delays and long silent gaps during live conversations • Degraded real-time experience due to retries and backoff behavior These issues are directly impacting the usability of our voice assistant in production-like scenarios. To better manage and scale our system, we would appreciate your guidance on the following: • What are the available usage tiers, and how can we identify which tier we are currently on? • What are the exact limits for: • Requests per minute (RPM) • Tokens per minute (TPM) • Are these limits enforced per API key, per session, or globally? • Is it possible to monitor our real-time usage against these limits?

Hey muratsk, Please share your org id

Hey! Confirmed, it's TPM, not RPM. Quick answers: • Check your tier + request an increase at console.anthropic.com → Limits • Limits are per API key • 429 headers show your current usage in real time Top fixes for voice: trim conversation history to last N turns + enable streaming. Happy to help you implement this want to connect? @muratsk

Oh thats great i never enter before console anthropic, im checking