Releases·Tool Release
Show HN: Nibia Fabric – Turn everyday computers into a shared AI compute cluster
A developer launched Nibia Fabric, an open-source system that pools CPU and RAM across macOS, Linux, and Windows machines to run LLM inference as a shared cluster. Built on llama.cpp, it offers capacity-aware scheduling, adaptive memory reservation, persistent tensor caching, and an OpenAI-compatible API. The alpha is validated on three physical machines, letting practitioners repurpose everyday hardware for distributed inference.
1 engagement1 sourceMon, Sep 28, 2026, 06:16 PM
AI Overview
- Nibia Fabric pools CPU and RAM across macOS, Linux, and Windows machines for distributed LLM inference, building on llama.cpp
- The open-source system adds capacity-aware scheduling, adaptive memory reservation, persistent tensor caching, and an OpenAI-compatible API
- The developer validated the alpha on three physical machines with a 30B-class Qwen model and GPT-OSS 20B
Related
Products·building
RAMDeck: pool RAM across old PCs over LAN to run 120B AI models locally
RAMDeck is a source-available tool and hardware hub that pools RAM from multiple old PCs over Ethernet to run large local LLMs. It exploits the sequential processing of transformer layers: weights load once at startup and only a small activation vector (~650 KB/token) crosses the network during generation. This lets users run models like a 120B-parameter LLM without expensive enterprise GPUs.
RAMDeck is a source-available tool and hardware hub that pools RAM from multiple old PCs over Ethernet to run large local LLMs. It exploits the sequential processing of transformer layers: weights load once at startup and only a small activation vector (~650 KB/token) crosses the network during generation. This lets users run models like a 120B-parameter LLM without expensive enterprise GPUs.
5 engagement1 sourceredditFri, Oct 16, 2026, 08:33 PM
Products·launched
Heterogeneous home cluster pooling RAM across laptop, PC, and Mac to run a 13B model
An orchestration layer on top of llama.cpp's ggml-rpc backend that lets mixed home machines (CPU-only laptop, CUDA PC, Metal Mac) share memory to serve a 13B LLM. It handles node discovery, memory-aware sharding, and failover, exposing an OpenAI-compatible API. Aimed at people who want to run larger models locally without a single big GPU.
An orchestration layer on top of llama.cpp's ggml-rpc backend that lets mixed home machines (CPU-only laptop, CUDA PC, Metal Mac) share memory to serve a 13B LLM. It handles node discovery, memory-aware sharding, and failover, exposing an OpenAI-compatible API. Aimed at people who want to run larger models locally without a single big GPU.
8 engagement1 sourceredditMon, Sep 7, 2026, 04:13 AM
Social·Community
User plans RTX 3060 12GB + RTX 5060 Ti 16GB for 28GB VRAM in llama.cpp
A user asks whether adding a used RTX 3060 12GB to an RTX 5060 Ti 16GB will work well in llama.cpp for running models like Qwen 3.6 27B at high quants. The setup would total 28GB VRAM across two GPUs with different PCIe generations.
A user asks whether adding a used RTX 3060 12GB to an RTX 5060 Ti 16GB will work well in llama.cpp for running models like Qwen 3.6 27B at high quants. The setup would total 28GB VRAM across two GPUs with different PCIe generations.
73 engagement2 sourcesredditTue, Aug 11, 2026, 01:33 PM
Products·building
P2P network for running local LLMs across multiple machines
DaiHive is a peer-to-peer network that splits large language models across multiple machines, allowing users to run models too big for a single computer. It supports private team mode for isolated groups, targeting users with multiple idle machines who want to run large models locally.
DaiHive is a peer-to-peer network that splits large language models across multiple machines, allowing users to run models too big for a single computer. It supports private team mode for isolated groups, targeting users with multiple idle machines who want to run large models locally.
2 engagement1 sourceredditFri, Jul 17, 2026, 09:36 AM
Social·Community
Users explore distributed LLM inference across multiple used phones
A Reddit user proposes running a distributed LLM across two used Poco F5 Pro phones to avoid the cost of a Mac or GPU, citing high RAM prices and cheap used phone memory. The plan involves rooting the phones, running Ubuntu in a chroot, and using llama.cpp's RPC support over WiFi.
A Reddit user proposes running a distributed LLM across two used Poco F5 Pro phones to avoid the cost of a Mac or GPU, citing high RAM prices and cheap used phone memory. The plan involves rooting the phones, running Ubuntu in a chroot, and using llama.cpp's RPC support over WiFi.
4 engagement1 sourceredditWed, Aug 26, 2026, 08:25 AM