This calculator models independent, uniformly random UUID v4 values with 122 random bits. It does not model truncated UUIDs, faulty randomness, v5, ULID, or timestamp-based UUIDs.
How the estimate works
With N = 2^122 and n generated identifiers, the birthday approximation is p ≈ 1 − exp(−n(n−1)/(2N)). We use a numerically stable calculation so very small probabilities remain visible. Counts at large scales are approximate because browser numbers have finite precision.
The 50% threshold is about 2.715 × 10^18 UUIDs. At one billion per day that is about 7.43 million years. A collection has many pairs: the probability that one pair matches is not the risk across the whole collection.
Frequently asked questions
- Does this guarantee my UUIDs will not collide?
- No. This is a probability model, not a guarantee. Defective randomness, truncation, and application bugs can invalidate its assumptions.
- Can I use this calculation for UUID v7?
- No. UUID v7 includes a timestamp and may use counters. Its collision model depends on the implementation and the number generated within the relevant timestamp interval.