Normal worker prompt tokens
Backend preview count for the untouched user prompt before the master prompt is added.
---- words
Token Cutter 500same model, two paths
One prompt in. Two workers out. Same model. Same master prompt.
The normal path sends your raw prompt straight through. The TK5 path strips filler words first, then sends the shorter version through the same model with the same instruction stack.
Backend-driven prompt metrics
Shared master prompt
TK5 stop-word stripping
Prompt input
Type once, then send the same request through both workers. Preview token charts refresh as you type.
82 / 220 wordsReady to sendLive token preview
TK5 worker prompt tokens
Backend preview count for the trimmed prompt that goes through the TK5 path.
---- words
Preview estimate for the full prompt until you send it.
Preview estimate for the trimmed prompt until you send it.
Appears after the live model call.
Appears after the live model call.
Words removed: --
Preview savings: -- tokens (--)
Live latency appears after you send the prompt.
Normal worker response
Full prompt sent directly to grok-4-fast-non-reasoning with the shared master prompt.
Input
--
Output
--
Total
--
Latency
--
The normal worker answer will appear here after the prompt is sent through the raw path.
TK5 worker response
Trimmed prompt sent to the same model with the same shared master prompt.
Input
--
Output
--
Total
--
Latency
--
The TK5 worker answer will appear here after the stop-word-stripped path finishes.