From the source

Apple Silicon and macOS VMs: 11–16× Faster LLM Inference with llama.cpp

Metal capability shim speeds llama.cpp in macOS VMs

Research paper

Image published with the post by Cua — Image from cua.ai

Cua released a research Metal capability shim that intercepts GPU capability queries inside macOS VMs, allowing llama.cpp to select newer Metal kernels.

On an M1 Ultra, TinyLlama 1.1B prompt processing improved 11.08× and token generation 16.36× versus a stock VM.

The shim is released under a permissive license for reproduction and further testing.

From the source
On an M1 Ultra, TinyLlama 1.1B running through llama.cpp processed prompts 11.08× faster and generated tokens 16.36× faster than the same workload in the same stock VM.
cua.ai

Most covered

OpenAI GPT-6 and Intelligent UI for everyone
SiliconANGLE: “With GPT-6, ChatGPT’s responses are much more visually appealing and interactive”+31 more →31 sources ▴
NVIDIA NVIDIA, Microsoft Kick Off a New Beginning for Windows PCs With RTX Spark and AI Agents
The Verge: “Windows and Surface event: how to watch and what to expect”+28 more →28 sources ▴
Google DeepMind We're making it easier to identify AI-generated content globally.
SiliconANGLE: “Google debuts SynthID Detector tool for flagging AI-generated content, but it’s far from perfect”+27 more →27 sources ▴
Anthropic Introducing Claude Haiku 5.5
Android Authority: “Claude Haiku 5.5 is here: The fastest, most affordable Anthropic model yet”+20 more →20 sources ▴
Google DeepMind EmbeddingGemma 2: an open, lightweight multimodal embedding model
SiliconANGLE: “Google expands EmbeddingGemma beyond text to images, audio and video”+9 more →9 sources ▴

Other stories

Other stories

Other stories

Apple Silicon and macOS VMs: 11–16× Faster LLM Inference with llama.cpp

From the source

Other stories

Blending the best of computer and browser use, without a Chrome extension

Computer-Use 2.0 at AI Engineer World's Fair

Evaluating Gemini 3.5 Flash on Computer-Use with Cua-Bench

Inside Linux computer-use: AT-SPI, XTEST, and background agents

Inside Windows computer-use: synthetic cursors for background agents

Inside macOS window internals: how SkyLight enables multi-cursor background agents