Independent researcher Jinho Jang published a cross-scale mechanistic study of safety training in Mixture of Experts reasoning models, reporting 40+ novel findings from 200+ controlled experiments acr…
Ferrox, a pure-Rust GGUF inference engine, now runs mixture-of-experts (MoE) prefill on Apple Metal 2.4x faster, reaching 1402 tok/s on OLMoE-1B-7B and closing the gap to llama.cpp from 2.62x behind t…
Ferrox v0.9.1, a pure-Rust GGUF inference engine, shipped today with MoE prefill on Apple Metal 2.4x faster, closing the gap to llama.cpp from 2.62x behind to 1.11x on OLMoE-1B-7B. The update also imp…
Salvatore Sanfilippo, creator of Redis, released h3.c, a native C inference engine for MiniMax's open-source H3 video model, rendering video with synchronized stereo audio on Apple Silicon in about 75…
Oleksii Vasylenko released g6k-rs, a research-oriented Rust library for lattice reduction, enumeration, and sieving on CPUs, Apple Metal, and NVIDIA CUDA, available on GitHub. The library includes flo…
Sovereign-Metal, a zero-dependency Python-to-Metal GPGPU advection engine and local transformer pipeline, has been released on GitHub under an MIT License. The engine enables zero-copy unified memory …
Mozilla AI released Llamafile 0.10.4, introducing Transcribefile built on the Transcribe.cpp speech-to-text inference library that supports 16+ TTS models. The update also includes improvements to Vul…
ZML launched LLMD, a free inference server that runs LLaMA, Gemma, Qwen, and Mistral models on NVIDIA, AMD, Google TPU, Intel, and Apple hardware from a single Docker image. Built in Zig and compiled …
A team rewrote a subset of the TinyGrad deep learning framework in Lean 4, creating TGrad, which outperforms the original in 4 of 5 benchmarks by up to 3x. The project demonstrates Lean 4's viability …
A researcher using nanoGPT on a MacBook Pro faces dependency failures when moving to a GPU cluster, prompting a solution using Nix and Flox to create reproducible, cross-platform runtime environments …
Developer Anil Madhavapeddy released ocaml-deepseek, an OCaml library that integrates DeepSeek's open-weight LLM directly into applications via a native inference engine, enabling local, dependency-fr…