NVIDIA RTX 5090 and DGX Spark Update Adds LTX 2.5 and Open AI Models

NVIDIA updated its DGX and RTX platforms to support the latest open AI models, which matters because developers can now run newer generative AI tools directly on existing hardware without waiting for new chips. This update expands the software library available to AI practitioners and gamers who use RTX GPUs for creative workloads. Buyers with RTX 5090 or RTX PRO systems get immediate access to optimized model support.

Performance gains include 26x faster animation generation on RTX 5090

The update targets NVIDIA DGX systems and GeForce RTX graphics cards, including the DGX Spark, DGX Station, GeForce RTX 5090, and RTX PRO series. NVIDIA stated that these systems now support recent open AI models such as Nemotron 3.5 and LTX 2.5. The company also highlighted support for Meta Muse Glimmer, Cosmos 3 Edge, MiniMax-H3, and Wan-Animate-2. This broad compatibility ensures that users can deploy a wide range of open-source models across different hardware tiers.

• Meta Muse Glimmer Performance: 200+ tokens per second on GeForce RTX 5090
• LTX-2.5 Performance Improvement: Up to 2x performance improvement and 40% memory reduction
• Wan-Animate-2 Performance vs Apple M3 Ultra: Up to 26x higher on GeForce RTX 5090, 16x higher on RTX PRO 5000 Blackwell
• LTX-2.5 GPU Optimization: 20% performance improvement and 40% memory reduction with RTX PRO 6000 96GB

Performance gains vary significantly depending on the specific model and GPU used. Meta Muse Glimmer, a 30-billion parameter model, runs at over 200 tokens per second on the GeForce RTX 5090. The LTX 2.5 video generation model sees up to a 2x performance improvement and a 40% reduction in memory usage on RTX GPUs. When using the RTX PRO 6000 96GB GPU, LTX 2.5 delivers a 20% performance boost and cuts memory usage by 40%. Wan-Animate-2 runs up to 26 times faster than Apple M3 Ultra on the RTX 5090 and 16 times faster on the RTX PRO 5000 Blackwell.

DGX Spark receives new software features including the NVIDIA Sync Cluster Assistant and native ARM64 Linux support for Google Chrome. The Sync Cluster Assistant allows users to connect multiple DGX Spark units into a high-speed cluster for workload distribution and system monitoring. NVIDIA also added a Sync Resource Monitor to track CPU and GPU usage in real time. These tools simplify the management of small-scale AI clusters for developers and researchers.

We looked at NVIDIA Brings Day-1 DiffusionGemma Support to in our earlier Nvidia coverage. This update builds on previous software enhancements for the RTX and DGX ecosystem. The changes focus on software optimization rather than new hardware releases.

NVIDIA confirmed that these updates are available for DGX and RTX systems. Users can now run Nemotron 3.5, LTX 2.5, and other open AI models with improved performance on supported hardware.