Prompt / Output Diff
Paste two prompt versions, or two model outputs, and see exactly what changed. Deletions are struck through in red, insertions highlighted in green, with change stats and a similarity score.
0
Words added
0
Words removed
0
Words unchanged
-
Similarity
Diff (A → B)
The highlighted diff will appear here.
Inputs are very large, so the changed middle section is shown as one removed block and one added block instead of a fine-grained alignment.
How it works: the two texts are tokenized (by word or by line, per the toggle above), then aligned with a longest-common-subsequence diff, the same family of algorithm used by git. The similarity score is 2 x matched / (length A + length B), so 100% means identical token sequences. In word mode, runs of whitespace are treated as single separators. Typical uses: comparing two iterations of a system prompt, checking what a model changed between two generations, or verifying that an edit only touched what you intended. Everything runs locally in your browser; nothing is uploaded.