Constellation Gate AI - Prompt injection and token savings - #1 in benchmarks
Point your AI agent at Gate. Inherit prompt-injection defense, secret scanning, and a verifiable audit trail. Gate includes prompt compression and caching so you can reduce token usage by 20-40% without changing model output. In AI security benchmarks, Gate ranks #1 across 16 public prompt-injection datasets. Keep your Claude or ChatGPT subscriptions and route them through Gate, or choose from 100+ models pay-as-you-go. Setup is easy: no code changes with our desktop app.
Replies
"#1 across 16 public prompt-injection datasets" is a strong claim - is that benchmark run by you internally or is there a third party leaderboard I can check that number against? also the 20-40% token savings "without changing model output" - does the compression step ever get caught out by a prompt that relies on exact wording (like a few-shot example)
Constellation Gate AI
@omri_ben_shoham1Â The benchmarks were run by our team and we published the complete methodology so that anyone can replicate the results. You can check out the blog post or arXiv report if you want to dig into the specifics.
The prompt compression never changes prompt wording - it saves tokens by removing whitespace and invisible tokens, and reformatting inefficient or duplicate tool results. There are specific strategies for different tools that are designed to optimize for token count without modifying the information available to the model.
appreciate the detailed answer, makes sense. curious about one edge case though - if a tool result already has whitespace stripped/minified (like some APIs return compact json by default), does the compression step just no-op on those or is there overhead either way
Macaly
prompt injection defense + token savings in one proxy is such a smart combo 🔒 zero code changes is the selling point imo
Constellation Gate AI
@petrkovacik Thank you!