13m
This is your 3rd spark 😱 how much is that in your local? Here we can’t find it.
4
Did you try with multiple DGX Sparks yet? My frieng bought one and has been very disappointed of the performance, memory bandwidth throttles is to just being a small demo of DGX
5
57m
3 can't be used for TP though... you want it for more concurrent sessions and more kv cache?
7
Fernando Diaz retweeted
After I complete my Local LLM series on @NVIDIAAI DGX Spark. Iw ill be writing 7 days of @GeminiApp!
Are you excited?
The Local LLM series has been doing good rounds on the internet! next is about the inference engines, LFG! I hope you like the animations ;)
1
2
13
403
Congrats to the @MiniMax_AI team on the release of MiniMax M3, a long-context multimodal model for text, image, and video reasoning. 🙌
Try it today with our free GPU-accelerated endpoint on build.nvidia.com.
Details: nvda.ws/4v4BWhD
Jun 12
MiniMax M3, Open-Weight, Now On Hugging Face , with only ~428B parameters and ~23B activated parameters
Weights:
huggingface.co/MiniMaxAI/Min…
MiniMax Sparse Attention:
huggingface.co/papers/2606.1…
51
116
1,314
134,347
@NVIDIAAI I'm developing a system to run local AI. I was wondering if I could get a DGX Spark to figure out the real concurrency you can obtain from that machine
16
Useful eval: compositional failure, not just FPS. When a requested behavior is missing or ambiguous, does the primitive refuse, recover, or choose a plausible wrong motion? For robotics, report contact-rich success under perturbation recovery latency.
4
also I just realized NVDA hosts free opensource models like k2.6 and m3 and glm5.1
low key waiting for k2.7 to drop in build.nvidia.com @NVIDIAAI
11
One open model. 350,000 motion clips. 15,000 FPS.
MotionBricks from NVIDIA Research runs real-time character animation at scale, without hand-crafted transitions or fine-tuning. And yes, it works for robotics too.
#SIGGRAPH2026 paper, demos code: nvlabs.github.io/motionbrick…
36
129
1,089
99,159