The Desk-Side Revolution: NVIDIA’s DGX Spark Update Delivers 2.5× AI Speed Boost

NVIDIA announced a major update to its DGX Spark desktop AI development platform at CES 2026, delivering performance gains of up to 2.5× through software optimisation alone and introducing a new suite of tools and workflows for Agentic AI development. The update also enables DGX Spark to operate in tandem with the latest RTX 5090 GPU, significantly speeding up 3D content creation pipelines.

Seven new Playbooks cover inference, fine-tuning and other workflows, with DGX Spark now supporting PyTorch fine-tuning across two linked DGX Spark systems for memory-intensive workloads. According to official benchmarks, performance on the Qwen-235B model has increased by more than 2.5×, while Stable Diffusion 3.5 Large and PyTorch fine-tuning workloads surpassed 2× gains, enabling faster inference and training on the same hardware.

The release also introduces Nsight Copilot, an offline AI assistant for CUDA programming, and MacBook Pro acceleration over a local network of up to 8×, reducing a 4K video generation task in ComfyUI from eight minutes to one. In addition, NVIDIA announced a partnership with Hugging Face to power Reachy Mini, and said the NVIDIA AI Enterprise software suite will add DGX Spark support by the end of January.