- Planned software update adds external accelerator role to Nvidia’s DGX Spark
- MacBook Pro users can push heavy AI processing to the external system
- Performance and tooling changes focus on local open source AI
At CES 2026, Nvidia revealed it is planning a software update for DGX Spark which will significantly extend the device’s capabilities.
Powered by the GB10 Grace Blackwell Superchip, Nvidia’s tiny powerhouse combines CPU and GPU cores with 128GB of unified memory, letting users load and run large language models locally without relying on cloud infrastructure.
The first reviews of the Spark, while universally positive, pointed out limitations with the software. This is something Nvidia is looking to address. A core part of the change will be expanded support for open-source AI frameworks and models.
Good news for MacBook Pro users
The move will be a software-only change, with no new hardware involved – and for organizations which rely on open tools, the changes will reduce custom setup work and help keep systems running in the same way as models and frameworks evolve.
The update to the mini PC will add support for tools such as PyTorch, vLLM, SGLang, llama.cpp, and LlamaIndex, as well as models from Qwen, Meta, Stability, and Wan.
Nvidia claims users can expect up to 2.5x performance gains compared with Spark at launch, driven primarily by TensorRT-LLM updates, tighter quantization, and decoding improvements.
One example Nvidia shared involved Qwen-235B, which more than doubles throughput when moving from FP8 to NVFP4 with speculative decoding. Other workloads, including Qwen3-30B and Stable Diffusion 3.5 Large, reportedly show smaller gains.
The update also introduces DGX Spark playbooks that bundle tools, models, and setup guides into reusable workflows. These are designed to run on-prem without rebuilding entire environments.
One interesting demonstration paired a MacBook Pro with DGX Spark for AI video generation. Nvidia showed a 4K pipeline that took eight minutes to complete on the laptop and roughly one minute when compute-heavy steps were offload to the Spark.
The approach keeps creative tools on the laptop while Spark handles the heavy processing, moving AI video work closer to interactive use instead of long batch runs.
DGX Spark can also act as a background processor for 3D workflows, generating assets while creatives continue working on their main systems.
A local Nsight Copilot is included, allowing CUDA assistance without sending code or data to the cloud.
Taken together, the planned update will move DGX Spark from a standalone developer system into a flexible on-prem AI node able to support laptops, workstations, and edge deployments.
Via StorageReview
TechRadar will be extensively covering this year’s CES, and will bring you all of the big announcements as they happen. Head over to our CES 2026 news page for the latest stories and our hands-on verdicts on everything from wireless TVs and foldable displays to new phones, laptops, smart home gadgets, and the latest in AI. You can also ask us a question about the show in our CES 2026 live Q&A and we’ll do our best to answer it.
And don’t forget to follow us on TikTok and WhatsApp for the latest from the CES show floor!
https://cdn.mos.cms.futurecdn.net/MrgqW79aMthXDsiDL9wdyj-1400-80.png
Source link
waynewilliams@onmail.com (Wayne Williams)
