Why does the free tier say 429?
Rate limits (RPM, TPM, RPD) · Builder · 5 min · tokens, llm · after Tokens
A 429 means a rate limit is full. Free APIs meter you several ways at once: requests per minute, tokens per minute, requests per day. Whichever fills first rejects calls until it frees up.
1 Guess first
2 Look
Three meters, one key
1 / 4requests / minute0 / 30tokens / minute0 / 8,000requests / day0 / 1,000Three meters, one key. Nothing sent yet.
illustrative requests of 1,500 tokens each
limits: Groq Free plan, openai/gpt-oss-20b, console.groq.com/docs/rate-limits (checked 2026-10-21)
A free key is metered three ways at once: requests per minute, tokens per minute, requests per day.
Requests, tokens and days are separate meters; the first one to fill answers 429, and the response tells you how long to wait.
RPM is requests per minute, TPM tokens per minute (your prompt plus the answer), RPD requests per day. Groq’s free plan lists all three for each model; OpenRouter’s free models list per-minute and per-day request limits. A long prompt costs one request but thousands of tokens, which is why TPM often fills first.
4 You are the model
You are the API server. For each incoming request, look at what this key has used and answer 200 OK or 429 Too Many Requests.
request 1 of 6 · you are the Groq server
This minute: 4 requests served, 7,600 tokens
New request: 600 tokens
Groq Free plan (gpt-oss-20b) and OpenRouter :free limits, checked 2026-10-21
Check every meter: one full meter is enough for a 429.5 Check
6 Say it back
7 Mission · 5 min
Create a free Groq key, send one chat request with curl -i, and find the x-ratelimit-remaining-tokens and x-ratelimit-remaining-requests headers in the reply. Then open your limits page in the console and match the numbers.
Three quick ones
Drawn at random from this lesson's pool; different next time.
Done?
Sources
- https://console.groq.com/docs/rate-limits (Free Plan Limits tab, read in Chrome, checked 2026-10-21; no update date on the page): openai/gpt-oss-20b 30 RPM, 1K RPD, 8K TPM, 200K TPD; “You can hit any limit type depending on which threshold you reach first.”; 429 + retry-after in seconds; “Rate limits apply at the organization level, not individual users.”; header example retry-after: 2
- https://openrouter.ai/docs/api-reference/limits (checked 2026-10-21; no update date on the page): :free variants 20 requests/minute, 50/day with less than 10 credits purchased, 1000/day with at least 10; 429 from OpenRouter or the upstream provider; “Retry with exponential backoff … Honor the Retry-After header when present.”
- Request sizes in the picture (1,500 tokens each) are illustrative, labelled as such
Published Oct 21, 2026 · updated Oct 21, 2026