A 0.57 MB router model cuts LLM costs by 68%
Negaaaa7 · reddit · 2026-07-20
The author claims they cut LLM costs by 68% with Prompt Compass, a tiny 0.57 MB model that handles four classifications in a single CPU call of about 5 ms.
What it does
It combines four tasks that are often split across multiple models and network hops:
- prompt routing
- PII scanning
- jailbreak detection
- local/cloud lane selection
Reported results
- 2,022 held-out real prompts
- 82% overall accuracy on the hard 4-way task
- 87.5% jailbreak recall on lmsys/toxic-chat
- under 5% false blocks on genuine prompts
- 100% recall on 15 obfuscated PII attacks in their benchmark comparison
Where it fails
- about 1 in 5 prompts are misrouted, mostly from local to cloud, which hurts cost rather than safety
- persona prompts like “act as...” often hit the jailbreak lane and are flagged for review
- it does not catch slow-burn multi-turn attacks because it only sees one prompt at a time
Integration
The tool ships as:
- an on-device npm SDK
- a hosted API
- a VS Code / Cursor extension
The extension is a thin client to the hosted API, while the SDK runs locally. It cannot intercept built-in Copilot/Cursor chat. The extension is also available on Open VSX and has a free tier.
Related event: Tiny 0.57MB Router Slashes LLM Costs(2 posts)→
More from coding & agent
- SF 'Code Engineering' event picks up where 1,000-person harness engineering meetup left off — LangChain · 2026-10-27
- Prompt turns Claude Code into an iPhone backup forensics tool exposing 30-day 'deleted' photo copies — alex_verem · 2026-10-27
- Prompt turns Claude Code into a forensics tool finding deleted iPhone photos in backups — alex_verem · 2026-10-27
- Phones won't be mobile IDEs but control planes for agent fleets, argues Yuchen Jin — Yuchenj_UW · 2026-10-27
- FrontierCS 2 launches on Kaggle: 48 tasks make agents beat published CS papers — MeganRisdal · 2026-10-27
- Stop remaking old games with AI: a call for more ambitious agentic software — rschu · 2026-10-27