Unity offers the Inference Engine (formerly Unity Sentis), a tool that lets game developers embed and run custom AI models directly on a player's device — inside any game or app built with Unity — without requiring an internet connection.

Originally launched as Unity Sentis in open beta in November 2023, the package was renamed Unity Inference Engine when Unity AI launched with Unity 6.2 in August 2025, then reverted to the display name Sentis as of package version 2.4. It is a neural network inference library that allows developers to import trained AI models in ONNX format and run them locally in real-time on any platform Unity supports, using the device's CPU or GPU. Typical uses include AI-driven character behavior, real-time object detection, procedural animation, and other gameplay features that benefit from on-device computation. Unlike the Unity AI Assistant and Generators, Sentis runs locally and does not consume Unity Points.