https://localllmguide.com/ https://localllmguide.com/images/seo/home/preview-16x9.webp What can your computer run? Enter your RAM, graphics memory and first task to get a starting app and model. https://localllmguide.com/guides https://localllmguide.com/images/seo/guides/preview-16x9.webp Install your first local model Choose an app, follow the instructions for your operating system and check your first reply. https://localllmguide.com/software https://localllmguide.com/images/seo/software/preview-16x9.webp Pick local AI software by the job it performs. Choose desktop chat, an API for another program, or a workspace for a collection of PDFs. https://localllmguide.com/models https://localllmguide.com/images/seo/models/preview-16x9.webp Find a local model Compare the exact downloads, context limits and license terms of models that suit your task. https://localllmguide.com/hardware https://localllmguide.com/images/seo/hardware/preview-16x9.webp How much model can your hardware hold? Pick the profile with your memory configuration. Check system RAM and dedicated GPU memory separately. https://localllmguide.com/compare https://localllmguide.com/images/seo/compare/preview-16x9.webp Which app, model or setup should you choose? Compare two apps side by side, or choose a guide for your memory size and task. https://localllmguide.com/troubleshooting https://localllmguide.com/images/seo/troubleshooting/preview-16x9.webp Troubleshoot local AI Start with what failed: the download, loading the model, getting a reply or finding a passage in a PDF. https://localllmguide.com/tools/ram-vram-calculator https://localllmguide.com/images/seo/ram-vram-calculator/preview-16x9.webp Will this model fit at your context length? See how much memory the weights and conversation need after reserving room for other apps. https://localllmguide.com/guides/what-local-llm-can-my-computer-run https://localllmguide.com/images/seo/what-local-llm-can-my-computer-run/preview-16x9.webp What local LLM can my computer run? Find your installed RAM and dedicated VRAM, then compare one model at the context you intend to use. https://localllmguide.com/compare/ollama-vs-lm-studio https://localllmguide.com/images/seo/ollama-vs-lm-studio/preview-16x9.webp Ollama vs LM Studio Use LM Studio to browse and load model files in a desktop interface. Try Ollama to manage named models from a terminal or another app. https://localllmguide.com/guides/local-llm-ram-vram-requirements https://localllmguide.com/images/seo/local-llm-ram-vram-requirements/preview-16x9.webp Which memory does a local model use? On a PC with a dedicated GPU, compare the model with VRAM and check system RAM separately. On a unified-memory machine, allow for other apps in the same pool. https://localllmguide.com/compare/best-local-ai-setup-for-mac https://localllmguide.com/images/seo/best-local-ai-setup-for-mac/preview-16x9.webp Run a local model on a Mac Check your chip and unified memory. Compare the selected model with the memory still available while your everyday apps are open. https://localllmguide.com/compare/best-local-ai-setup-for-windows https://localllmguide.com/images/seo/best-local-ai-setup-for-windows/preview-16x9.webp Run local AI on Windows: app choice and setup Follow either LM Studio or Ollama from download to your first local conversation. https://localllmguide.com/guides/install-ollama https://localllmguide.com/images/seo/install-ollama/preview-16x9.webp How to install Ollama Install Ollama, run a downloaded tag and use ollama ps to see where the model loaded. https://localllmguide.com/guides/install-lm-studio https://localllmguide.com/images/seo/install-lm-studio/preview-16x9.webp Download and install LM Studio Find a GGUF in Discover, select its compression and load the downloaded file into a new chat. https://localllmguide.com/guides/open-webui-with-ollama https://localllmguide.com/images/seo/open-webui-with-ollama/preview-16x9.webp Use Open WebUI with Ollama Get Ollama working first. Then connect the browser interface to the local model service. https://localllmguide.com/guides/chat-with-pdfs-locally https://localllmguide.com/images/seo/chat-with-pdfs-locally/preview-16x9.webp How to chat with PDFs locally Try a question with a known answer on one searchable PDF. If the cited passage is missing, check extraction and retrieval separately. https://localllmguide.com/news/ollama-decision-models-nimble-setup https://localllmguide.com/images/seo/ollama-decision-models-nimble-setup/preview-16x9.webp Sort tickets locally with Nimble Use Ollama's decision API, define the allowed answers, and inspect the result before automating. https://localllmguide.com/news/qwen38-splash-bionic-mac-setup https://localllmguide.com/images/seo/qwen38-splash-bionic-mac-setup/preview-16x9.webp Run Qwen3.8 with Splash Check the Mac requirements for Splash and select the Qwen3.8-27B package built for that engine in Bionic. https://localllmguide.com/local-ai-privacy https://localllmguide.com/images/seo/local-ai-privacy/preview-16x9.webp Where data can go in a local AI setup Find where the app sends prompts and PDF text, where it saves chats, and which folders sync. https://localllmguide.com/guides/local-ai-linux-setup https://localllmguide.com/images/seo/local-ai-linux-setup/preview-16x9.webp Install a local model on Linux Start the runner, open a fresh terminal, send a local request and check whether the model loaded on the CPU or GPU. https://localllmguide.com/guides/use-local-ai-offline https://localllmguide.com/images/seo/use-local-ai-offline/preview-16x9.webp Use local AI without internet Download the runner and model, disconnect, then open a new chat and try the document search you need. https://localllmguide.com/guides/local-models-for-coding https://localllmguide.com/images/seo/local-models-for-coding/preview-16x9.webp Choose a local coding model Give the model a function that sorts when it should preserve order. Run the supplied checks on its proposed fix. https://localllmguide.com/guides/local-models-for-writing https://localllmguide.com/images/seo/local-models-for-writing/preview-16x9.webp Choose a local writing model Give two local models the same draft. Compare whether they preserve the names, deadlines and qualifications you supplied. https://localllmguide.com/guides/local-ai-6gb-vram https://localllmguide.com/images/seo/local-ai-6gb-vram/preview-16x9.webp Run local AI with 6 GB of GPU memory Try Qwen3 4B and check the PROCESSOR column in ollama ps while it is loaded. https://localllmguide.com/guides/local-ai-cpu-only https://localllmguide.com/images/seo/local-ai-cpu-only/preview-16x9.webp Run local AI without a dedicated GPU Time a reply from a small model twice: once after loading and again while it is still in memory. https://localllmguide.com/guides/local-ai-unlimited-tokens https://localllmguide.com/images/seo/local-ai-unlimited-tokens/preview-16x9.webp What unlimited tokens really means Local AI still has context, memory and speed limits even without a per-token bill. https://localllmguide.com/guides/local-ai-context-windows https://localllmguide.com/images/seo/local-ai-context-windows/preview-16x9.webp Choose a useful context window Budget instructions, chat history, document excerpts and the reply inside the context your app actually allocated. https://localllmguide.com/guides/local-ai-quantization https://localllmguide.com/images/seo/local-ai-quantization/preview-16x9.webp Local LLM Quantization (Model Compression) Explained Compare two compressions of the same model on one prompt. Record file size and any answer detail lost at the smaller size. https://localllmguide.com/guides/choose-gpu-for-local-ai https://localllmguide.com/images/seo/choose-gpu-for-local-ai/preview-16x9.webp Choose a GPU for local AI Choose a target model and context. Check whether the file and its conversation memory fit the card you plan to buy. https://localllmguide.com/guides/move-local-model-storage https://localllmguide.com/images/seo/move-local-model-storage/preview-16x9.webp Move your local model downloads Point the app to the new drive, reopen it and load a model from there before deleting the old copy. https://localllmguide.com/compare/gguf-vs-mlx https://localllmguide.com/images/seo/gguf-vs-mlx/preview-16x9.webp GGUF or MLX for your local setup Compare a GGUF file for llama.cpp with a complete MLX model package for a supported Apple silicon runner. https://localllmguide.com/compare/ollama-vs-llama-cpp https://localllmguide.com/images/seo/ollama-vs-llama-cpp/preview-16x9.webp Ollama or llama.cpp Choose between managing named downloads with Ollama and selecting files and GPU layers directly in llama.cpp. https://localllmguide.com/compare/lm-studio-vs-jan https://localllmguide.com/images/seo/lm-studio-vs-jan/preview-16x9.webp LM Studio or Jan Check which app loads your chosen file, where it stores chats, and which provider receives each request. https://localllmguide.com/troubleshooting/ollama-command-not-found https://localllmguide.com/images/seo/ollama-command-not-found/preview-16x9.webp Fix Ollama command not found Open a fresh terminal and locate the executable before reinstalling. Then check whether the Ollama service responds. https://localllmguide.com/troubleshooting/lm-studio-model-will-not-load https://localllmguide.com/images/seo/lm-studio-model-will-not-load/preview-16x9.webp Fix a model that will not load Distinguish an incomplete download, an unsupported model and insufficient memory. https://localllmguide.com/troubleshooting/poor-local-ai-answers https://localllmguide.com/images/seo/poor-local-ai-answers/preview-16x9.webp Improve poor local AI answers Give the model a question with a known answer. For a PDF, check whether the supporting passage reached the model. https://localllmguide.com/tools/token-counter https://localllmguide.com/images/seo/token-counter/preview-16x9.webp Count text and PDF tokens locally Count the readable text in a PDF with a Qwen3 tokenizer and see how much space remains for the reply. https://localllmguide.com/tools/browser-ai https://localllmguide.com/images/seo/browser-ai/preview-16x9.webp Try a small model in your browser Load Qwen3 0.6B through WebGPU. Closing the tab clears the chat; model files may stay cached. https://localllmguide.com/news/liquid-ai-lfm3-small-llm-laptop https://localllmguide.com/images/seo/liquid-ai-lfm3-small-llm-laptop/preview-16x9.webp Liquid AI LFM3 for local tasks A 1.7B text model with a 1.10 GB Q4 download. Check conversation memory and your own hardware. https://localllmguide.com/news/mistral-large-4-run-locally-hardware https://localllmguide.com/images/seo/mistral-large-4-run-locally-hardware/preview-16x9.webp Can you run Mistral Large 4 locally? The GGUF publisher lists a 298 GB download and a 512 GB unified-memory example for that conversion. Check the file before budgeting a workstation. https://localllmguide.com/news/drex-max-decision-model-mac https://localllmguide.com/images/seo/drex-max-decision-model-mac/preview-16x9.webp Run Drex Max on a Mac A 27B model that returns probabilities instead of text. Text-only on Macs with 24 GB or more. https://localllmguide.com/news/nvidia-rtx-spark-laptops-local-llm-memory https://localllmguide.com/images/seo/nvidia-rtx-spark-laptops-local-llm-memory/preview-16x9.webp RTX Spark laptops for local LLMs Up to 128 GB of unified memory on Windows. Check the memory size and your apps' Windows on ARM versions. https://localllmguide.com/news/mimo-v26-distill-qwen-9b-ollama-16gb https://localllmguide.com/images/seo/mimo-v26-distill-qwen-9b-ollama-16gb/preview-16x9.webp MiMo-V2.6 9B in Ollama Xiaomi's coding model has a 6.6 GB download, reads text and images, and uses the MIT license. https://localllmguide.com/news/lm-studio-0428-bionic-120-omlx-update https://localllmguide.com/images/seo/lm-studio-0428-bionic-120-omlx-update/preview-16x9.webp LM Studio 0.4.28 and Bionic 1.2.0 A fix for Mac updates that reopened the old version, a new oMLX engine, and on-device Read Aloud. https://localllmguide.com/news/open-webui-013-update-tabbed-workspace https://localllmguide.com/images/seo/open-webui-013-update-tabbed-workspace/preview-16x9.webp Open WebUI 0.13: back up, then update The update changes the database and adds a tabbed workspace. Install 0.13.2 or later. https://localllmguide.com/news/ripple-65b-preview-ollama-memory-requirements https://localllmguide.com/images/seo/ripple-65b-preview-ollama-memory-requirements/preview-16x9.webp Ripple-65B Preview: how much memory? The smallest Ollama download is 39 GB. A Mac with 64 GB of unified memory is the realistic minimum. https://localllmguide.com/news/does-local-ai-stay-local-privacy-audit https://localllmguide.com/images/seo/does-local-ai-stay-local-privacy-audit/preview-16x9.webp Does local AI stay local? A model can run on your computer while document indexing, tools or sync send information elsewhere. Trace one harmless file. https://localllmguide.com/news/gemma-4-local-mac-windows-guide https://localllmguide.com/images/seo/gemma-4-local-mac-windows-guide/preview-16x9.webp Gemma 4 on Mac and Windows Pick a Gemma 4 variant for the memory and inputs you need. Check the exact download and any image component first. https://localllmguide.com/news/gpt-oss-20b-vs-qwen3-8b-vs-gemma-4-12b-best-local-llm https://localllmguide.com/images/seo/gpt-oss-20b-vs-qwen3-8b-vs-gemma-4-12b-best-local-llm/preview-16x9.webp gpt-oss 20B, Qwen3 8B or Gemma 4 12B? Compare gpt-oss 20B and Qwen3 8B on a text task. For image input, check the Gemma 4 12B package and the app that loads it. https://localllmguide.com/news/liquid-ai-lfm25-small-local-models https://localllmguide.com/images/seo/liquid-ai-lfm25-small-local-models/preview-16x9.webp Liquid AI LFM2.5: choose by job One model extracts fields, one routes a tool call and one ranks passages. Each returns a different kind of output. https://localllmguide.com/news/lm-studio-bionic-local-vs-cloud https://localllmguide.com/images/seo/lm-studio-bionic-local-vs-cloud/preview-16x9.webp LM Studio Bionic: what is local? Bionic can use your Mac, a linked machine or hosted compute. Check the selected provider before giving an agent private files. https://localllmguide.com/news/ollama-032-agent-local-boundary https://localllmguide.com/images/seo/ollama-032-agent-local-boundary/preview-16x9.webp Ollama 0.32 adds an agent The bare command opens an agent. For a local chat, run a downloaded model by name and check cloud settings first. https://localllmguide.com/news/open-webui-011-security-subagents-sharing https://localllmguide.com/images/seo/open-webui-011-security-subagents-sharing/preview-16x9.webp Open WebUI 0.11: what to review Check the security advisory and patched version first. Then review who can share chats and what background agents can reach.