Palette v2.1 Release

Hello,

We’re excited to announce the release of Palette v2.1, launching April 16th, 2026!

This release introduces the Model Executor framework for flexible, high-performance inference across diverse model types and input modalities, expands GenAI runtime support with LFM2 and Qwen VL.


:rocket: Release Highlights

Model Executor

  • Supports 10X and Non-10X models with multiple input types: Video/Image, Audio, and Text
  • Sequential and async multi-model execution
  • GStreamer, C++ API, and Python API integration with Generic Pre-Proc and GenericBoxDecoder support

GenAI & Platform

  • New GenAI runtime support for LFM2 and Qwen VL on the SiMa.ai Modalix platform
  • Linux Kernel 6.18 and Ubuntu 24.04 support for SDK and PCIe host driver
  • 16 GB SOM support, low power mode, and secure boot
  • Automatic Graph Surgery for simplified model optimization

Hardware & Sensors

  • Expanded MIPI camera sensor support: IMX568, IMX477, IMX678, AR234, and OG05C10
  • ROS rover testing with RTABMAP and Cartographer SLAM, NAV2 navigation, and USB Camera pipelines.

:link: Get Started


— The SiMa.ai Team

2 Likes