Safety · The Decoder ·

Mistral's open model Shieldstral matches much larger safety models at a fraction of the size

Mistral's open model Shieldstral matches much larger safety models at a fraction of the size

Mistral introduced Shieldstral, a 3B open model that checks AI inputs and outputs for safety violations through natural-language yes-or-no questions. It reportedly matches models up to seven times larger on some benchmarks, supports runtime-defined criteria, and can run locally.

Read the full story at The Decoder →