Máy tính giới hạn tốc độ LLM miễn phí

Câu hỏi thường gặp

Phần này hiện bằng tiếng Anh.

How is "requests per day" computed for a tier?

From the published limits only: the smallest of requests per day, requests per minute times 1,440 (the per-minute cap sustained all day, a ceiling rather than a promise), tokens per minute times 1,440 divided by your average request size, and, where the vendor publishes them, tokens per day divided by your average request size and requests per month divided by 30. The row says which bound limits you.

Why is a tier listed under "limits not published"?

Its vendor publishes no requests-per-minute, requests-per-day or tokens-per-minute figure for the free tier on a page we can cite, so there is nothing to compute from. aifree.dev never guesses a number; the offer page links the vendor page and notes.

How does the combination work?

Greedily: the largest tier that falls short goes first, then the next largest from a different provider, up to three providers, until the sum covers your need. Keys from one vendor are never combined because they usually share one quota. It is a spread of load, not a single key.

What limits are not covered?

Daily token caps and monthly call caps enter the arithmetic once they are recorded (shown under the model name), but some tiers still carry them only in the notes column, and monthly token caps or credit balances are not modelled. A tier with such a cap in its notes may be overstated; read its notes before relying on it.

Is the result the same for everyone?

Yes. The result is a pure function of the catalog and the two numbers in the URL, so a shared link shows the same answer until the next monthly re-check updates a limit.