Browse Comments — Clean (de-noised)
Close reading of the corpus at each pipeline stage: raw → clean → relevant → coded.
40
comments matched
· page 1 of 2
Awesome innovation 💕❤️💕Showcase @ Computex 2026. See ya 🥰
and?????
Corriente already has all llms runing on one dgx spark. We can compress 100TB down to 28MB
Lewin Wanzer behold this box here... the internet! 🎁
On-premise AI for critical infrastructure just got a lot more compelling. No cloud dependency, no latency, no data sovereignty risk. This is the hardware story that makes what we’re building at Utilyst inevitable. 💧
This will make for a compelling addition with some of the work we're doing in the AI appliance space. Empowering organizations with meaningful AI workflows on-prem can lead to meaningful reduction in the majority of their cloud spend.
How much it costs? In-house solution is for poor))) tell me how much it costs?
Thats 200B model on Q4 Quantisation not FP16
I need one of this OMG
I'm not sure about how practical this mini PC would be to train any AI models, especially in terms of cooling during time-consuming AI model training processes. It might only survive as a thin client.
Cuda is the problem
Daniel Uhlemann my mac m1 ultra has similar specs. Might not be as fast but i had it for 5 years b
Right things to say is it has probably enough ssd or memory to fit the model. That’s all. Nothing else about perf, tco can be claimed.
The pace of AI hardware innovation is becoming extraordinary. Running large-scale models locally on a compact system would have seemed unrealistic just a few years ago. Developments like this could significantly expand access to advanced AI by reducing dependence on massive cloud infrastructure and centralized compute.
This could become a very important turning point for edge AI. As models become smaller, more efficient, and increasingly optimized for local hardware, AI may gradually move from massive centralized data centers toward personal devices and local inference systems. The long-term implications for privacy, latency, and AI accessibility could be enormous.
How does this compare to the NVIDIA DGX Spark ( and other GB10s ) and the Apple Mac Pro Ultra/Max ?
which run on the NVIDIA Graphics ? 😂
That’s what I need to use on Edge.
Good AMD is back
Future brains of ultra agile robotics systems