#localai — Public Fediverse posts

Local ai is finally being seen as a market.

😂

https://www.technology.org/2026/09/03/acer-rtx-spark-veriton-ri110-ai-desktops/

#localai #ai

Local ai is finally being seen as a market.

😂

https://www.technology.org/2026/09/03/acer-rtx-spark-veriton-ri110-ai-desktops/

#localai #ai

Local ai is finally being seen as a market.

😂

https://www.technology.org/2026/09/03/acer-rtx-spark-veriton-ri110-ai-desktops/

#localai #ai

Local ai is finally being seen as a market.

😂

https://www.technology.org/2026/09/03/acer-rtx-spark-veriton-ri110-ai-desktops/

#localai #ai

Local ai is finally being seen as a market.

😂

https://www.technology.org/2026/09/03/acer-rtx-spark-veriton-ri110-ai-desktops/

#localai #ai

I have a $2,500 Mac Mini with an M5 Pro chip coming in the next few weeks, and I'm looking forward to setting it up, but I also know what the next few purchases need to be because of it.  Thunderbolt 5 on a Mac allows Remote Direct Memory Access. RDMA lets one computer directly read and write memory on another computer without involving the CPU, and tools like Exo use Tensor Parallelism to split large AI models across multiple Mac Studios or Mac Minis, making separate computers act like one massive unified memory system.

If I understand this correctly, as long as I keep the number of computers even and connect them with Thunderbolt cables, I can build an AI cluster that pools unified memory.  So, when I add a second Mac Mini with 64 GB of memory, I will have 128 GB of memory, and when I add a third Mac Mini with 64 GB of memory and a Mac Studio with 64 GB of memory, I will be at 256 GB.

It might take half a decade, but I like having a solid plan.

#AI #LocalAI #Exo #RDMA

The author exhaustively analyzes the Qwen 3.8:27B model at different levels of quantization to see at what point accuracy drops below a minimal threshold.

#AI #LLM #Qwen #localai #localllm

https://kaitchup.substack.com/p/qwen38-27b-gguf-benchmark-q4-to-q1

The author exhaustively analyzes the Qwen 3.8:27B model at different levels of quantization to see at what point accuracy drops below a minimal threshold.

#AI #LLM #Qwen #localai #localllm

https://kaitchup.substack.com/p/qwen38-27b-gguf-benchmark-q4-to-q1

The author exhaustively analyzes the Qwen 3.8:27B model at different levels of quantization to see at what point accuracy drops below a minimal threshold.

#AI #LLM #Qwen #localai #localllm

https://kaitchup.substack.com/p/qwen38-27b-gguf-benchmark-q4-to-q1

The author exhaustively analyzes the Qwen 3.8:27B model at different levels of quantization to see at what point accuracy drops below a minimal threshold.

#AI #LLM #Qwen #localai #localllm

https://kaitchup.substack.com/p/qwen38-27b-gguf-benchmark-q4-to-q1

The author exhaustively analyzes the Qwen 3.8:27B model at different levels of quantization to see at what point accuracy drops below a minimal threshold.

#AI #LLM #Qwen #localai #localllm

https://kaitchup.substack.com/p/qwen38-27b-gguf-benchmark-q4-to-q1

@LokiTheCat it allows people to share their docs into the federation and then other people add stuff and it becomes the corpus of the sector - osint/comp intel, indexes, specialized insular industry info but all these people are working together for the greater good - you can run bigger models and they can be trained on docs specific to your sector so you get better answers #semantic tags #metadata #filtered lists #bloomberg termial killer #rag pipelines #trends #real time dashboards #distributed inference #localai