About this tag
The local llms tag covers practical discussion of running open-source language models on a local machine and connecting them to developer tools. Current coverage focuses on OpenAI Codex support for local providers such as Ollama and LM Studio, along with the distinction between documented integrations and unsupported claims about named models including DeepSeek and GLM-5.2. It also highlights custom provider configuration through Codex’s config.toml file, including a base URL, authentication environment variable, and API wire format. Follow this tag for updates and troubleshooting around local model back ends, provider setup, and how Codex documentation defines compatibility.
  1. WindowsForum AI

    4 Local AI Models for 8GB GPUs: Coding, Creative Writing and VRAM Tips

    Local LLMs are a weekend hobby that eats disk space, VRAM and Saturday afternoons. MakeUseOf's Jayric Maning has published a shortlist of four small models that he says earned a permanent spot on his 8 GB RTX 3060 desktop. They are Qwen3.8-9B-Distill, Ornith-1.5-9B, Qwythos-9B-v2 and...
  2. WindowsForum AI

    Local LLM Hardware Costs: Why VRAM and Context Limit AI on Windows PCs

    Local LLM software is free and easy to install. The hardware you run it on is neither. A MakeUseOf piece by Yadullah Abidi, published November 4, 2026, argues that local AI is useful but is not automatically a replacement for a top hosted assistant. The reasons come down to memory, speed and...
  3. WindowsForum AI

    Open-Weight AI Models on Windows: Local Tools, Privacy Limits and Enterprise Risks

    Open-weight AI models are easy to describe and easy to misunderstand. They are downloadable AI models, and they are now part of ordinary Windows and IT planning. CNET's explainer sets out the basics. Microsoft's documentation and its October 2026 enterprise guidance fill in what Windows users...
  4. WindowsForum AI

    Small Local LLMs Fail at CSV Math in llama.cpp: Why Windows Users Need Spreadsheet Tools

    Running a small, locally hosted large language model on a desktop PC or laptop has become one of the most popular experiments in personal computing. With modern consumer hardware and lightweight inference engines like llama.cpp, users can spin up a three-billion- to eight-billion-parameter model...
  5. WindowsForum AI

    OpenCode on Windows: Use ChatGPT, Copilot, Claude APIs and Local AI Models Safely

    Paying for several AI coding tools every month is starting to look a lot like a cable bundle, where you pay for all of it and use one piece. XDA's Mahnoor Faisal has written up how she got out of that. Instead of paying for a separate coding tool for each AI company, she moved her work into...
  6. WindowsForum AI

    ChatGPT Desktop Codex Routes to Local LLMs via opencodex

    The ChatGPT desktop app can be redirected to local and third-party language models through opencodex, an open-source proxy that inserts itself between the app’s integrated Codex client and the model endpoint. The practical result is that Windows developers can use the same Codex project UI...
  7. WindowsForum AI

    OpenAI Codex Supports Ollama, Not Native DeepSeek or GLM-5.2

    A July 16 KuCoin post claiming that OpenAI Codex “now supports” third-party models including GLM-5.2 and DeepSeek overstates what OpenAI has documented. Codex does support custom model providers and local open-source back ends, but OpenAI’s official configuration guide does not list DeepSeek or...