Category: Local LLM
Underdog Woof: Run the 32 MB Text-to-Speech Model Locally (Specs, Setup, Licence)
Underdog Woof is a 32 MB English text-to-speech model that runs on one CPU thread with no GPU. Specs, vendor benchmarks, setup commands, Python code, licence and how…
Intel Arc GPU for Local AI (2026): B580, B60, B70 with Ollama, LM Studio, llama.cpp and ComfyUI
Is Intel Arc good for AI in 2026? Every Arc card from B570 to the 32 GB Pro B70 with prices, what models fit, measured B580 speeds, the…
Nemotron 3.5 Transcribe: Run NVIDIA’s Open 40-Language Speech Model Locally (GGUF, VRAM, Setup)
NVIDIA’s open Nemotron 3.5 Transcribe turns up to an hour of audio into text in 40 languages with speakers, timestamps and translation. Specs, benchmarks, which GGUF fits your…
Ripple-65B Preview: Specs, Benchmarks, Price and How to Run It Locally
Ripple-65B Preview from Ricursive: a 65B dense Apache 2.0 coding model. Benchmarks, $0.20/$0.60 API price, GGUF sizes, llama.cpp, DGX Spark setup and fixes.
Inkling 2 Small: Thinking Machines’ 594B Model (Specs, Benchmarks, Pricing, Open Weights)
Inkling 2 Small from Thinking Machines Lab: 594B/17B MoE with image input, AA Index 31, $0.30/$1.00 Tinker pricing, how to call it, hardware for the coming Apache 2.0…
How to Run MiniMax-M3.1-Flash Locally (2026): RAM Needed, GGUF Quants and llama.cpp Setup
MiniMax-M3.1-Flash open weights are out: 428B MoE, 23B active. See how much RAM and VRAM each GGUF quant needs, llama.cpp build and run commands, offload settings, vLLM servers…
vLLM vs llama.cpp vs Ollama (2026): Benchmarks, Hardware and Which to Use
vLLM, llama.cpp and Ollama compared: single-user and 32-user benchmarks, hardware and OS support, model formats, RAM offload, setup commands, OpenAI API and which one fits you.
How Much RAM Do You Need for Local AI? LLMs, ComfyUI and MoE Models (2026)
Clear RAM numbers for local AI in 2026: 32 GB vs 64 GB vs 128 GB for Ollama, llama.cpp, LM Studio and ComfyUI image and video, MoE offload…
Unlimited OCR 2: Run Baidu’s 3B Long-PDF OCR Model Locally (VRAM, GGUF, SGLang, Benchmarks)
Baidu’s MIT-licensed Unlimited OCR 2 parses 100-300 page PDFs in one pass. Benchmarks, VRAM needs, and step-by-step local setup with transformers, SGLang and patched llama.cpp GGUF, plus fixes.
Open WebUI + Ollama Setup Guide (2026): Docker, Python, Connect Models and Fix Errors
Install Open WebUI with Docker or Python, connect it to Ollama, turn on document chat, web search and tools, and fix the common connection errors.