Edge0 announces 145B model and Kinetic local inference engine

First seen Nov 10 Updated Nov 13, 05:39 UTC Heat 46 high evidence

Edge0 announced a 145B model called Edge0-145B, pitched as frontier capability on a personal computer. The Kinetic repo is an open-source Rust engine that runs GGUF models on CPUs and NVIDIA, AMD and Intel GPUs. Its README claims 4.5x to 4.6x faster generation than llama.cpp on one laptop.

What to do: Install Kinetic and run a small GGUF model such as LFM3 1.7B to compare speed on your own hardware.
Who should care: Developers running local models who use GGUF files and llama.cpp.

Why it matters: Kinetic is Apache-2.0 and reads existing GGUF files, so developers can try it on CPU or NVIDIA, AMD and Intel GPUs from one binary.

Key facts

Kinetic license: Apache-2.0
CPU test, LFM3 1.7B: 128.1 t/s vs 28.1 t/s for llama.cpp
GPU test, LFM3 1.7B: 1033.7 t/s vs 229.7 t/s
Test hardware: Ryzen 5 6600H, RTX 3060 Laptop 6 GB; Compared against llama.cpp b12001
Repo stars: 86

Context

The README benchmarks Kinetic against llama.cpp, which uses the same GGUF format.

Timeline

Coverage

Nov 10, 11:03 UTCNov 10
Veloresearch/kinetic: The open engine for running AI on your own computer. Fast, exact, open source.github.com · GitHub · 86 stars

Discussion

Nov 12, 23:06 UTC6h ago
Edge0-145B: Frontier AI Capabilities on Your Own Computeredge0.ai · HN · 150 pts · 38 comments

Evidence

High evidence: the claim comes from an official source (the company or project itself), or at least two independent sites reported it.

Details written Nov 13, 05:38 UTC from the sources listed above.