Teknozip – Nvidia is making significant strides in the realm of local artificial intelligence with the introduction of its latest software tool, Underdog. Designed to complement the recently launched RTX Spark series, this application aims to simplify the complex process of managing and running large language models (LLMs) directly on personal hardware.
The RTX Spark devices, which arrived on the market in mid-October, are high-performance, Arm-based System-on-Chip (SoC) machines running Windows. These laptops are engineered specifically to handle the heavy computational demands of local AI. Major hardware manufacturers including ASUS, Dell, HP, Lenovo, Microsoft, and MSI have already released versions of these devices, with Acer expected to follow suit.
To unlock the full potential of these machines, Nvidia has released the Underdog app. The software utilises a specialised inference engine known as the Husky model, which allows users to run sophisticated AI models locally. A primary advantage of this approach is privacy; because the processing occurs entirely on the device without requiring cloud connectivity, sensitive user data remains securely on the laptop.
Also Read
Performance enthusiasts will also benefit from significant technical upgrades. Nvidia has confirmed new speculative decoding support via MTP for the Qwen 3.8 Flash Next model. When paired with an RTX Pro 6000 Blackwell, this enhancement can accelerate end-to-end agentic tasks by up to 1.4 times.
Furthermore, Nvidia is ensuring that users stay at the cutting edge through "day-zero support." This initiative ensures that several prominent models are compatible with Nvidia systems immediately upon their release, eliminating the usual waiting period for driver or software updates.
The following models are included in this immediate support rollout:
| Model Category | Supported Models |
|---|---|
| Language Models | Nemotron Bolt E4B 4B, GLM, DeepSeek, Qwen, MiniMax |
| Diffusion Models | FLUX 3 Klein |
By bridging the gap between complex AI architecture and user-friendly management, Nvidia is positioning its RTX Spark ecosystem as a premier choice for developers and privacy-conscious AI enthusiasts alike.