Articles tagged "ローカルAI"

EmbeddingGemma 2 runs in 191MB RAM (text only) and about 1.5GB BF16 for all modalities, no GPU. Memory table, setup, local RAG. October 2026.

Whistle is Cactus's 16.9MB open speech-to-text model for CPU, Apache 2.0. It supports 7 languages but not Japanese. Speed vs Whisper base, setup, best uses.

Cloudflare Clef and Clef-flash: open-weight decision models with a Jev-compatible API. Apache 2.0, 41GB minimum VRAM, $0.24/M. Compared with AWS Strands Decider 2B and Jev.

Irodori-TTS-v4-Large is a 3.29B Japanese-only TTS with voice cloning, Voice Design and emoji control. Setup, benchmarks vs v4.1-Small, Gemma license notes.

Ollaya is an open-source local runtime for decision models: state and typed questions in, calibrated probabilities out in ms, via Ollama-style pull/run/serve.

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.

A guide to Speedstu/CUDA-for-AMD-Windows, which hit 135 points on Hacker News. Covers how this reproducible ZLUDA + AMD HIP/ROCm stack runs CUDA-targeted Windows apps on AMD GPUs, validated hardware, installation, what works and what doesn't, and how it compares to plain ZLUDA, ROCm on Linux, DirectML, and Vulkan-based llama.cpp.

A guide to debpalash/VoiceStudio (formerly OmniVoice-Studio), which surged on GitHub Trending in September 2026. Covers how it runs voice cloning, voice design, video dubbing, dictation, transcription, and audiobook creation entirely on local hardware, its 16 TTS / 11 ASR engine lineup, installation, local API/MCP support, and how it compares to ElevenLabs and other tools.

Magnitude is an open-source local inference server that profiles your hardware, then recommends, downloads, tunes, and runs the best model for your machine. This guide covers setup, how it plugs into Claude Code and other agents, and how it differs from Ollama, LM Studio, llama.cpp, and vLLM, as of September 2026.

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.

MiniMax-Music3 is an open-weight text-to-music model that generates up to 5-minute tracks as 32kHz/16-bit stereo WAV. The reference setup targets 24GB+ VRAM, with CPU offloading fitting ~22GB and layer-by-layer streaming reaching down to 8GB cards. As of August 2026, this article lays out the confirmed hardware requirements, license terms, and the paid API's new-user sunset starting August 20, 2026.

LTX-2.5 (22B DiT, Gemma 4 encoder) is an open video+audio model out Aug 11, 2026. VRAM: 16-80GB; files: 35.9-71.4GB by quant. See table and GPU picks.

Comparing local LLM engines: Ollama and llama.cpp solo, MLX on Apple Silicon, vLLM for concurrent serving, TensorRT-LLM for NVIDIA, LM Studio for GUI trials.

TypeWhisper is a local-first speech-to-text app. Press a hotkey, speak, text inserts at cursor. Free open-source core for macOS/Windows, no telemetry.

A 2026 comparison of Ollama and LM Studio, the two leading local LLM runners: CLI vs GUI, install and usage, API, OS support, and model formats, with guidance on which to pick.

Gemma 4 needs 5GB VRAM (E2B/E4B), 16GB (26B MoE), or 24-62GB (31B Dense) depending on quantization. Requirements by model: RTX 3060 to H100, Apple Silicon M1-M4, CPU-only operation, RAM sizing, and budget builds. Updated July 2026.

A practical migration guide to building a Claude replacement using Qwen 3.5-9B (refreshed Q2 2026). Apache 2.0 license, 262K context, runs on 16GB RAM. Complete coverage from Ollama setup to API migration, prompt conversion, cost comparison, and current limitations as of April 2026.

By combining Ollama and OpenClaw, you can build AI agents on Mac mini that dynamically switch between multiple LLMs. This article provides detailed practical steps from Ollama installation to model management, OpenClaw integration configuration, and performance comparison. We introduce how to build a local AI infrastructure that can be adopted by SMBs and startups, especially in Shinagawa, Minato, Shibuya, Setagaya, Meguro, and Ota wards.

A comprehensive guide to building high-performance AI agents with Qwen3.5-9B using Mac mini M4 and OpenClaw. Covers hardware requirements, LINE/Slack/Discord integration, and performance benchmarks.

Comprehensive guide to Qwen3.5-9B: Ollama setup instructions, hybrid Gated DeltaNet + Sparse MoE architecture, 262K context window, GPQA 81.7 and IFBench 76.5 (beating GPT-5.2's 75.4), comparison with GPT-4o-mini and Claude Haiku, and practical business use cases. Runs on just 5GB RAM.

A complete step-by-step guide to installing Qwen3.5-9B locally on Mac, Windows, and Linux. Covers setup via Ollama, llama.cpp, and vLLM, along with quantization options (GGUF Q4/Q5/Q8), GPU acceleration (CUDA/Metal), Docker deployment, API server configuration, performance tuning, and troubleshooting. Practical instructions for developers and IT administrators looking to run SLMs on-premises.

A thorough TCO comparison between running Qwen3.5-9B as a local AI and using cloud APIs. Covers hardware costs, electricity, maintenance, break-even analysis, and ROI calculation frameworks for SMBs in Shinagawa, Minato, and Shibuya.

Gartner has named Domain-Specific Language Models a top strategic technology trend for 2026. Small Language Models (SLMs) are transforming AI adoption for SMBs with lower costs, higher accuracy for specific tasks, and zero data leakage risk. This guide covers benefits, leading models, practical use cases, and step-by-step adoption.