News
Latest news and updates on local artificial intelligence
9 news items in Hardware & accelerators
Intel and AMD merge CPU optimizations into oneDNN for PyTorch
October 20, 2026AMD merges its ZenDNN fork into oneDNN and joins the UXL board; Intel brings ACE extensions for matrix multiplication. oneDNN becomes the default x86 library in PyTorch: one software path to optimize for local AI on CPU, regardless of vendor.
llama.cpp b11717: Hadamard for Q1_0 and Q2_0 quantizations on Snapdragon NPU
October 18, 2026Build b11717 adds Hadamard MUL_MAT and GET_ROW operations for Q1_0 and Q2_0 quantizations (flat and tiled) to the Hexagon backend (Snapdragon NPU), and reduces packet counts in the q8_1 quantizer. Qualcomm contribution: a further step toward mobile NPU Inference for the most aggressive quantizations.
From garage to professional maker lab: Raspberry Pi tells it in a free game
October 15, 2026Raspberry Pi launches a free game where you grow a tinker garage into a professional maker lab: the company's story told through progressive upgrades.
KNOD: in-kernel GPU offloading for network packet processing
October 10, 2026The open-source KNOD project lets AMD GPUs handle network packet processing without specialized hardware, with high performance and low power draw. A development update was shared at the Linux Plumbers Conference in Prague.
Linux 7.4: laptop and gaming handheld support improvements incoming
October 10, 2026The Linux 7.4 kernel merge window is receiving a batch of x86 platform driver improvements for laptops and gaming handhelds: many driver branches have been queued ahead of the next kernel version.
Boro: NVIDIA brings local AI to Linux kernel development
October 9, 2026NVIDIA is developing Boro, a Rust CLI that brings local AI to Linux kernel development: patch validation, builds, boot and tests, built for backporting. Led by Andrea Righi and presented at LPC 2026, it is Apache 2.0 and runs on local back-ends (Claude, OpenCode, Codex) or OpenAI-compatible servers.
vLLM v0.31.0rc3: randomized dummy inputs support in Model Runner V2
October 1, 2026The v0.31.0 release candidate for vLLM adds support for randomized dummy inputs in the new Model Runner V2 (PR #58411), a step toward more robust testing and profiling of the new execution engine.
AMDXDNA: Linux driver enhancements for Ryzen AI 300 NPU3
September 26, 2026A series of 21 patches for the open-source AMDXDNA driver extends NPU3 (Ryzen AI 300) support: kernel-mode command submission, full suspend/resume handling and Telemetry, paving the way for NPU-based local AI on Linux.
NVIDIA NemotronLabs VoiceChat 11B: real-time open voice conversation with tool calling
September 22, 2026NemotronLabs VoiceChat 11B is an end-to-end full-duplex voice model: audio understanding and generation in a single architecture, without the cascaded ASR→LLM→TTS stack. It is the first open full-duplex model to support tool calling while keeping the natural flow: ~450ms responses, interruptions, and natural turns. The code is on GitHub.