Token Cutter 5000

Beta testing for a new prompt-compression technology created by Justin Gund and Grok. During testing, both workers return formatted TLDR answers with a heading and bullet points, capped at 300 words or less, to keep token costs down while preserving the response outcome.

Backend-driven prompt metrics

Formatted TLDR mode

Side-by-side worker answers

Temp 0 • Max 512 • TLDR heading + bullets • 300-word cap

Prompt input

Type or copy/paste your prompt in here. 220 word limit. To keep comparisons and token cost down during testing, both workers return a formatted TLDR with a heading and bullet points, capped at 300 words or less, instead of a full response. Send all inquiries to justingund@level3id.com.

0 / 220 wordsAdd a promptLive token preview

Normal worker TLDR response

Full prompt sent to grok-4-fast-non-reasoning with the shared formatted TLDR prompt.

Input

Output

Total

Latency

Prompt --Completion --Finish after runFingerprint after run

The normal worker TLDR will appear here after the prompt is sent through the raw path.

TC5 worker TLDR response

Compressed prompt sent to the same model with the same formatted TLDR prompt.

Input

Output

Total

Latency

Prompt --Completion --Finish after runFingerprint after run

The TC5 worker TLDR will appear here after the compressed path finishes.

Normal Tokens In--

Preview estimate for the full prompt until you send it.

TC5 Tokens In--

Preview estimate for the compressed prompt until you send it.

Normal Tokens Out--

Appears after the live model call.

TC5 Tokens Out--

Appears after the live model call.

Formatted TLDR mode — both sides use the same model, same settings, and the same TLDR heading + bullet prompt capped at 300 words or less.

Preview savings: -- tokens (--)

Live latency appears after you send the prompt.

Compression may change response style slightly, but the testing goal is to preserve the same outcome.

Normal worker prompt tokens

Backend preview count for the untouched user prompt before it is sent to the model.

---- words

TC5 worker prompt tokens

Backend preview count for the compressed prompt that goes through the TC5 path.