Fastino Labs Open-Sources GLiGuard: A 300M-Parameter Safety Moderation Model That Matches or Exceeds the Accuracy of Models 23–90x Its Size

Fastino Labs open-sourced GLiGuard, a 300M-parameter safety moderation model for LLM guardrails.

It treats moderation as classification, checking multiple safety dimensions in a single forward pass.

The company says GLiGuard performs comparably to models 23 to 90 times larger.

It also runs up to 16 times faster, which could cut cost and latency for production AI apps and agents.