Nvidia's new Underdog app makes managing your local AI models even easier - XDA
Nvidia has released its new Underdog app to help users manage local AI models on RTX Spark devices, which launched October 16 across major OEMs.
Nvidia's new Underdog app simplifies managing local AI models on its recently launched RTX Spark devices. These powerful Arm-based System-on-Chip devices, running Windows, hit the market on October 16 and are available from major OEMs including ASUS, Dell, HP, Lenovo, Microsoft, and MSI, with Acer expected to join soon.
The Underdog application utilizes a Husky model-specific inference engine to run local AI models directly on users' systems. Unlike cloud-dependent solutions, this approach keeps all data securely on the laptop, addressing privacy concerns while enabling powerful local processing capabilities.
Performance Boost
Nvidia has enhanced performance with new speculative decoding support using MTP for Qwen 3.8 Flash Next. This technology enables end-to-end agentic tasks to run approximately 1.4x faster when operating on an RTX Pro 6000 Blackwell system, leveraging the llama.cpp framework.
new speculative decoding support with MTP for Qwen 3.8 Flash Next
Model Support
The app offers day-zero support for several new models including Nemotron Bolt E4B 4B, GLM, DeepSeek, Qwen, MiniMax, and FLUX 3 Klein diffusion model. This immediate compatibility means users can access these models as soon as they're released without waiting for additional support updates from Nvidia.
Related
