Originally published at: Open-Source AI Tool Upgrades Speed Up LLM and Diffusion Models on NVIDIA RTX PCs | NVIDIA Technical Blog
AI developer activity on PCs is exploding, driven by the rising quality of small language models (SLMs) and diffusion models, such as FLUX.2, GPT-OSS-20B, and Nemotron 3 Nano. At the same time, AI PC frameworks, including ComfyUI, llama.cpp, Ollama, and Unsloth are making functional advances, doubling in popularity over the past year as the number…
@annamalaic It’s great to see Nvidia working on popular community tools like ComfyUI to improve performance. Long may it continue!
I would like to bring an issue to your attention: due to the unusual architecture of the Nvidia DGX Spark, models loaded in ComfyUI use 2x the memory. As a result, the 128GB Spark runs out of memory once you load 64GB of models, reducing it from a great system to an overpriced one.
You can read more about this issue here: Buyers beware: DGX Spark limited to 64GB in ComfyUI
I hope you can bring it to the attention of the same Nvidia developers who helped with ComfyUI and Flux2 in your blog post.