Articles tagged "Apple Silicon"

Google AI Edge Gallery runs Gemma 4 12B (6.9GB) offline on a 16GB Apple Silicon Mac. Setup and benchmarks: 15 tok/s on a MacBook Air M4. Updated Sept 2026.

Released Sep 2026, Edge0-35B-A3B-preview 4-bit-quantizes Qwen3.5-MoE 35B-A3B and streams experts off SSD on demand, running at under 3GiB of active memory and roughly 15-18 tok/s on MLX. Here's how it works, what hardware you need, how to run it, and how it compares to llama.cpp mmap offload and Colibri.

As of Sep 2026, Perplexity's Hybrid Compute feature for the Mac Computer agent starts tasks in the cloud and hands off only the steps involving sensitive data to a local model (PPLX Qwen 3.8 27B) on Apple Silicon. This article explains, based on primary sources, how the local/cloud boundary is decided, the hardware requirements, and how it differs from Apple's Private Cloud Compute and fully local setups.

oMLX is an open-source LLM inference server built for Apple Silicon Macs, combining a tiered KV cache (RAM + SSD) with continuous batching and a menu-bar macOS app for management. As of August 2026 it sits near the top of GitHub Trending with roughly 20,000 stars. This guide covers its features, requirements, setup, and how it compares to Ollama, LM Studio, mlx-lm, and llama.cpp.

Qwen Scribe runs Qwen3-ASR on Apple Silicon via MLX for fully offline transcription and system-wide dictation — 3.4GB unified memory for the 1.7B model, 1.2GB for 0.6B. Covers requirements, setup, push-to-talk usage, SRT export, the three macOS permissions, and how it compares to Whisper-based tools.

**Apple Container** is **Apple's official Swift OSS for running Linux containers on macOS**, announced at WWDC 2025 ([GitHub: apple/container](https://github.com/apple/container) / [apple/containerization](https://github.com/apple/containerization) / [Apple Open Source](https://opensource.apple.com/projects/container/) / [WWDC25 session](https://developer.apple.com/videos/play/wwdc2025/346/)). **v1.0.0 shipped June 9, 2026** under Apache 2.0, **44.5k GitHub stars and 1.3k forks** at writing, 98% Swift, **Apple Silicon only**. **The defining design choice is its "one VM per container" architecture** — unlike Docker Desktop's shared-kernel VM, each container runs in its own lightweight VM for stronger security and resource isolation. Sub-second boot times, minimal root filesystem, default 1 GiB RAM and 4 CPUs per container, and **near-zero idle footprint when nothing is running**. **Tech stack**: macOS 26's Virtualization.framework + vmnet framework + XPC + launchd + Keychain. The control plane is container-apiserver / container-core-images / container-network-vmnet / container-runtime-linux. **OCI-compatible** with Docker Hub / GHCR; build with the BuildKit-based `container builder`. Cross-arch (arm64 / amd64), with x86 running under Rosetta. **Where it fits vs Docker Desktop**: Apple Container is strongest at **single-container runs, native isolation, and minimal idle cost**; Docker Desktop still wins on **Compose, ecosystem maturity, and multi-platform support**. **Docker Compose is not supported at v1.0.0**, memory ballooning is partial (released pages may not return to the host — heavy loads may require restarts), and these limits are explicit in the docs. **Requirements**: **Mac with Apple Silicon + macOS 26** (macOS 15 works with networking constraints; Intel Macs are fully unsupported). **Use cases**: local backend services, CI-style builds, cross-architecture image generation, data analysis via host-folder mounting, and untrusted-code isolation. It's also **an excellent companion for running [local LLMs](../columns/local-llm-landscape-2026-june-update) on M5 Macs** — Ollama / vLLM containers paired with Apple Container is a natural fit. The column closes with three inquiry funnels for Mac developer environment setup, container migration, and ongoing maintenance.

Gemma 4 E4B is Google's 4.5B parameter edge AI model released in April 2026. This guide covers local deployment on Apple Silicon and Raspberry Pi, multimodal features, quantization settings, and benchmark comparisons.