SiMa.ai has announced Palette 3.0, a major software release designed to improve the performance of Physical AI applications, including autonomous drones and robotics. The latest version builds on the company’s Agentic Palette software suite and introduces improved concurrent model execution, open-source capabilities and support for advanced generative AI workloads.
According to SiMa.ai, Palette 3.0 can increase model processing speed by up to 3.5 times while operating within a low-power budget. The release is designed to help developers run multiple AI models simultaneously for machine reasoning and real-world actions.
SiMa.ai Palette 3.0 Improves Concurrent AI Model Execution
One of the key upgrades in Palette 3.0 is optimised concurrent model execution. The platform allows multiple AI models and complex reasoning workloads to run simultaneously on a single Modalix MLSoC.
It also supports running separate Docker containers on-chip, enabling developers to execute different AI workloads within the same system. These include large language models (LLMs) and vision-language models (VLMs), while working within power and memory constraints.
The improved execution capabilities are designed to support Physical AI applications that require multiple models to process information and make decisions in real time.
Improved AI Processing for Autonomous Drones and Robotics
Palette 3.0 accelerates perception, depth and reasoning models used in autonomous drones and robots. SiMa.ai has also introduced a Model Browser that allows developers to explore the platform’s performance across key AI models.
The Model Browser includes performance measurements under real-world power constraints. The company says its platform also works with ecosystem partners to support unified sensor fusion, dense neural depth fields and closed-loop digital twin simulation.
These capabilities aim to address integration and validation challenges in autonomous system development.
Palette 3.0 Adds Open-Source GenAI Development Capabilities
SiMa.ai has introduced additional capabilities for generative AI application development. The software supports EAGLE3, DFlash and Gemma 4 MTP speculative decoding algorithms, which the company says can increase model tokens per second (TPS) by up to 3.5 times while maintaining a low-power operating profile.
Another key addition is support for running Mixture-of-Experts (MoE) models with large parameter counts. This includes OpenAI’s GPT-OSS-20B model on SiMa.ai’s Modalix MLSoC 16GB System-on-Module (SoM).
The company is also highlighting open-source code that developers can adapt for their applications. These additions build on its Agentic Palette software, which SiMa.ai says can reduce AI development timelines from months to days or hours.
SiMa.ai Model Browser and Performance Benchmarking
SiMa.ai’s Model Browser provides performance information for key models used in Physical AI applications, including autonomous drones and robotics.
Developers looking for detailed competitive benchmarking information can contact SiMa.ai directly at contact@sima.ai.
With Palette 3.0, the company aims to expand its Physical AI software capabilities by combining concurrent model execution, improved processing performance and support for more complex AI workloads within a low-power computing environment.
Leadership Comment
“The fundamental bottleneck in Physical AI today is the glaring mismatch between software that evolves monthly and legacy silicon that takes years to adapt. We are breaking that cycle,” said Krishna Rangasayee, founder and CEO of SiMa.ai. “Following our industry-first Agentic Palette announcement, Palette 3.0 delivers on our core thesis: hardware at the speed of software. By continuously maximizing superior performance-per-watt for any Physical AI model, we give customers the ability to deploy tomorrow’s breakthrough capabilities on today’s hardware, completely rewriting the roadmap for fully autonomous systems.”
Also read – Girikon.AI Introduces Real-Time Voice Translation Platform to Help Enterprises Overcome Language Barriers
Join us on LinkedIn for quick updates – FYI9 News LinkedIn


