# SenseAuto Unveils Sage, an On-Device Multimodal Agent Foundation Model

- Category: Flash (快讯)
- Published: 2026-04-22T11:52:00.000Z
- Source date: 2026-04-22
- HTML: https://www.88z.ai/en/news/flash-20260422-b7c5876c2dfeef0b
- Markdown: https://www.88z.ai/en/news/flash-20260422-b7c5876c2dfeef0b.md
- Other language: https://www.88z.ai/news/flash-20260422-b7c5876c2dfeef0b.md

SenseAuto has launched Sage, the industry's first complex agent foundation model for vehicles. Built on a MoE architecture, it achieved strong results in PinchBench benchmarks and is deployed on the NVIDIA Orin X platform.

## At a glance

- Sage is the industry's first complex agent foundation model for vehicles, using a MoE architecture.
- It achieved a 94% task completion rate in PinchBench and is deployed on the NVIDIA Orin X platform.

## The update

SenseAuto has officially launched Sage, an on-device multimodal agent foundation model. It is the industry's first foundation model designed for complex agents in vehicles. The model uses a Mixture of Experts (MoE) architecture with 32 billion total parameters and 3 billion activated parameters. In PinchBench benchmarks, Sage achieved a 94% task completion rate, outperforming leading global cloud-based models such as Claude-Opus-4.6 and GPT-5.4. Sage has been successfully deployed on the NVIDIA Orin X platform. Furthermore, SenseAuto plans to introduce Sage Box at the Beijing Auto Show, aiming to evolve smart cockpits into super agents and enhance in-vehicle interaction and processing capabilities.

## Sources

- [绝影智能SenseAuto](https://mp.weixin.qq.com/s/KaSuIXuwZPCe-wvirfUHMQ) (2026-04-22)

This article is an AI-assisted synthesis by 88Z AI based on the public sources below. Product claims and availability are attributed to their publishers. It does not represent independent testing or first-hand reporting.