
NVIDIA DGX Spark 64 GB, which facilitates local artificial intelligence development processes, is available at prices starting from $ 4,999.
AI-generated summary
NVIDIA is developing a new product line focused on running AI models on local hardware rather than in the cloud.
A critical move has come from NVIDIA for developers and researchers who want to run artificial intelligence models directly on their own hardware without cloud dependency. The company officially announced the NVIDIA DGX Spark 64 GB model, the new member of the series that it positions as a personal artificial intelligence supercomputer. Compact system based on Grace Blackwell architecture; It meets users with a price tag starting from $ 4,999 through leading manufacturer partners such as Acer, ASUS, Dell, Gigabyte, HP and MSI.
Grace Blackwell Power in a Compact Case: What Does DGX Spark 64 GB Offer?
The new DGX Spark configuration preserves the GB10 Grace Blackwell Superchip processor, DGX OS operating system and NVIDIA's comprehensive software ecosystem in the 128 GB upper model of the series. Thanks to its unified memory architecture, the device can run open source artificial intelligence models and autonomous agents with up to 100 billion parameters completely locally, without risking data privacy.
The system comes ready for developers' workflows right out of the box. NVIDIA Agent Toolkit directly supports popular operating environments such as Ollama, vLLM, LM Studio and PyTorch, as well as CUDA-X libraries and Nemotron models. It is also among the shared details that a platform-specific precompiled installer for the popular 3D modeling and animation software Blender will be released very soon.
Two Units in One System: Easy Clustering with NVIDIA Sync Cluster Assistant
One of the most striking aspects of DGX Spark 64 GB is its scalability. Thanks to the NVIDIA ConnectX-7 network adapter and QSFP connection integrated on the device, users can directly connect the two systems. NVIDIA Sync Cluster Assistant comes into play at this point and automatically takes care of network configuration and load distribution.
Memory Capacity: When two units are connected, the combined memory pool reaches 128 GB.
Model Support: Large-scale language models with up to 200 billion parameters can run on a single cluster.
Performance Increase: In Qwen 3.8 27B tests, two clustered systems offer up to 1.7x performance increase and twice the memory bandwidth compared to a single device.
Running Artificial Intelligence Models Comes Down to a Single Click
NVIDIA focuses on simplifying the software experience as well as hardware. NVIDIA Sync Model Launcher, which will be available towards the end of the month, reduces running models on a local device or cluster to a few clicks. Users can launch the Qwen3.8 27B model on DGX Spark with a single click from their laptops and can code directly from the browser thanks to OpenCode integration.
AI outlook — possibilities, not facts
Special precompiled installer for Blender will be released.
Very likely · Within weeks

Apple's new flagship iPhone 18 Pro Max was reviewed with the A20 Pro processor, variable aperture camera and under-screen Face ID technology. Offering a similar experience to previous generations in daily use, the device stands out with its camera and artificial intelligence performance.
TEKNOFEST, organized by the Ministry of Industry and Technology and the T3 Foundation, welcomes technology enthusiasts at Şanlıurfa GAP Airport. The festival, which takes place on an area of 191 thousand square meters, will continue with various events until October 4.

TEI General Manager Mahmut Faruk Akşit announced that the TS1400 engine developed for GÖKBEY has reached flight maturity and is aimed to receive a type certificate in 2027. In addition, the TF6000 engine is planned to be delivered to Baykar at the end of 2026 or early 2027.

Artificial intelligence model GPT-6 Astra deciphered an encrypted military letter sent to General Auguste de Marmont by the order of Napoleon Bonaparte in 1809 for the first time in 200 years.

ASELSAN exhibits air defense systems, unmanned vehicles and electronic warfare technologies at TEKNOFEST held in Şanlıurfa. The company also supports young talents by organizing scientific events for children with the 'Techno Adventure' area.

To facilitate the daily life of the visually impaired, Google has integrated the 'Guided Vision' feature, which offers real-time voice guidance, into the Gemini Live platform. The system, which works on devices with Android 9 and above, depicts the environment using the camera.