#Local AI

Google's Antigravity SDK can now run agents on local models with no internet connection. What changed, how to set it up and the limits Google states.

The 60 most-downloaded open text and vision-language models on Hugging Face, by 30-day downloads collected September 18, 2026. Qwen holds 27 rows; the caveats.

Edge0 streams a 35B mixture-of-experts model from SSD in under 3 GiB of RAM. Bonsai 2 shrinks a 27B model to 5.9 GB. Which to pick, and what to test first.

Test whether a local AI workflow depends on cloud models, search, embeddings or downloads, and record what works offline without overstating privacy guarantees.

Meta released Muse Glimmer 30B under a genuine Apache 2.0 license: a dense, multimodal model distilled from Muse Spark and aimed at local agent work.

Alibaba promised Qwen3.8-Max and a 27B sibling for the week of 10 August. Updated: Max weights landed 8 August under a bespoke licence. Run these checks.

Whisper-family engines, VoiceInk and Handy against the subscription apps: what open-source dictation matches today, and where paid tools still earn the fee.

AI PCs clear the 40+ TOPS Copilot+ bar, but can NPUs actually run local LLMs? A 2026 buyer's guide to NPU vs GPU, Phi Silica and what really matters.

Apple's MLX framework explained for developers: unified-memory zero-copy ops, mlx-lm generation and LoRA fine-tuning, and why M5 beats llama.cpp on Mac.

Match 2026's best open-weight coding models to your hardware: Qwen3-Coder-Next, Devstral 2, GLM-5.2 and DeepSeek V4 by VRAM, SWE-bench score and real speed.

When does a local AI rig beat cloud APIs? We run the 2026 break-even math: a $9K RTX PRO 6000 wins past ~26% daily use, but never beats commodity inference.

Run speech-to-text locally for $0 per minute with open Whisper. Compare whisper.cpp, faster-whisper, WhisperX and Parakeet, and find the cloud break-even.

A 3-9B small language model on your laptop can handle most agentic loop steps faster and cheaper than a frontier cloud model. Here's how to build SLM-first.

Compare the RTX 5090, RTX PRO 6000, DGX Spark, M5 Max and Mac Studio for local LLMs by price, VRAM, memory bandwidth and real tokens per second.

DGX Spark, M5 Max and RTX PRO 6000 Blackwell compared for local AI: memory, bandwidth, FP4 compute, real tokens per second and current 2026 pricing.

GGUF, AWQ, GPTQ, EXL2 and MLX quantization formats compared for 2026 — which fits your CPU, GPU or Mac, and how each affects speed and accuracy.

Run Flux, Stable Diffusion and ComfyUI locally in 2026 for $0-per-image, license-clean brand visuals — VRAM needs, model licenses and real speeds.

Apple raised Mac prices up to 33% on June 25, 2026 over a memory-chip shortage. We run the 3-year TCO of local AI hardware versus cloud subscriptions.

DiffusionGemma is Google's first open-weight text diffusion LLM: a 26B MoE under Apache 2.0 hitting 1,100+ tokens/sec on one H100. Where it wins and loses.

Gemma 4 12B processes text, image, audio, and video with no separate encoders, fitting in ~7GB at 4-bit. A guide to running private multimodal agents locally.

RTX Spark, DGX Station, Microsoft Scout, and Hermes Desktop all shipped in one week. Why the on-device agent shift reshapes cost, latency, and privacy.

Meta's Manus AI agent launches as a desktop app for local device integration. Direct AI agent access without cloud dependency. Full setup guide.