Musubi releases PolicyLM-1.7B, a lightweight decision model for real-time content moderation with open weights

Musubi, a startup focused on AI decision models, released PolicyLM-1.7B, a 1.7B-parameter model designed for real-time content moderation and released with open weights.

The model is intended to apply plain-English content policies to messages in under 50 milliseconds, matching the speed of AI classifiers used by social platforms but with the flexibility of a modern LLM.

Unlike traditional classifiers, PolicyLM-1.7B does not require retraining when policies change, enabling policy teams to iterate as needed without additional training cycles.

Musubi’s co-founder and chief AI officer, Filip Jankovic, stated that product teams need scalable, customizable ways to label content as platform volumes grow exponentially.

Decision models output outcome probabilities or binary judgments rather than text, allowing them to run faster and cheaper than large language models while retaining transformer architecture flexibility.

PolicyLM-1.7B is positioned as a self-hostable alternative to proprietary decision models, with Musubi highlighting its suitability for content moderation use cases.

The release follows increased industry interest in decision models since the emergence of TypeSafe AI’s Jev and subsequent releases from OpenAI and Amazon.