quelllm

Changes to the facts recorded here, not a live scan of every vendor update. Save this tool to follow updates in your account.

1. 9/14/2026

   Updated: best for, features, limitations, sources, summary, use cases, walkthrough

   See recorded changes

   bestFor

   Before: ["Users who want to run open-source LLMs locally on their PC or Mac with privacy and no cloud dependency.","French-speaking users seeking detailed guides and support for local AI model deployment and usage."]

   After: ["Users wanting to run open-source LLMs locally on Windows or Linux PCs and Macs with various GPUs and RAM configurations, prioritizing privacy and local computation.","Users needing local LLMs with large context windows for document-heavy applications and RAG workflows, with open licenses suitable for enterprise deployment."]

   features

   Before: ["Free configurator that analyzes your PC or Mac hardware to recommend compatible LLMs with installation instructions in French.","Comprehensive catalog of over 247 open-weight LLMs with VRAM requirements, speed estimates, and use cases, highlighting French models.","Educational kits sold for 24€ each or 49€ for all, covering specific AI use cases such as local RAG, local code copilots, local agents, and enterprise AI deployment."]

   After: ["Free configurator to check which LLMs run on your hardware with speed estimates and installation instructions in French.","Extensive catalog of over 220 open-weight LLMs with VRAM requirements, speed estimates, and use cases, including French models.","Educational guides and kits for various LLM use cases such as local chatbots, coding copilots, RAG, autonomous agents, and image generation.","Ranking and detailed information on open-source LLMs suitable for RAG with large context windows (≥ 32k tokens), quality synthesis, and permissive licenses.","Supports models with very large context windows up to 488,000 tokens, enabling processing of large document corpora."]

   limitations

   Before: ["Requires sufficient VRAM and compatible hardware to run large models locally, with detailed VRAM requirements provided per model."]

   After: ["RAG effectiveness depends on quality of document chunking and embedding; naive chunking can break tables and reduce comprehension.","Using embeddings trained on a different language than the documents reduces retrieval relevance significantly."]

   sources

   Before: [{"url":"https://quelllm.fr/","label":"LLM local : quel modèle tourne sur votre machine ? | QuelLLM.fr"}]

   After: [{"url":"https://quelllm.fr/","label":"LLM local : quel modèle tourne sur votre machine ? | QuelLLM.fr"},{"url":"https://quelllm.fr/meilleur-llm/rag","label":"Meilleur LLM local pour RAG en 2026 — contexte long, retrieval | QuelLLM.fr"},{"url":"https://quelllm.fr/guide/rag-local-introduction","label":"RAG local : introduction | QuelLLM.fr"}]

   summary

   Before: "QuelLLM.fr is a French platform providing a free configurator and guides for running over 220 open-source large language models (LLMs) locally on PC or Mac, focusing on privacy and local execution without cloud dependency. It offers detailed model catalogs, usage guides, and kits for various AI applications including RAG (Retrieval-Augmented Generation), local agents, and code copilots."

   After: "QuelLLM.fr is a French platform providing a free configurator and guides for running over 220 open-source large language models (LLMs) locally on PCs or Macs. It offers detailed information on compatible hardware, installation steps, and usage scenarios, emphasizing privacy and local computation without cloud dependency."

   useCases

   Before: ["Local RAG (Retrieval-Augmented Generation) to query personal documents like PDFs, notes, and emails without sending data to the cloud.","Local code assistance replacing GitHub Copilot and Cursor with a 100% local code copilot."]

   After: ["Running local LLMs for chat assistants, coding development, document retrieval and question answering (RAG), image and audio processing, reasoning, and autonomous agents.","Local Retrieval-Augmented Generation (RAG) for querying personal documents like PDFs, notes, and emails without sending data to the cloud."]

   walkthrough

   Before: []

   After: ["RAG involves two phases: indexing documents by chunking and embedding them into a vector database, then querying by embedding the question, retrieving relevant chunks, assembling a prompt, and generating a response with a local LLM.","Minimal RAG stack includes a local generation LLM (e.g., Qwen 3.5 9B), embedding model (e.g., multilingual-e5-large), vector database (e.g., ChromaDB), orchestration framework (e.g., LlamaIndex), and optionally a UI like Open WebUI or AnythingLLM.","Example minimal Python pipeline for RAG: load and chunk documents, configure embeddings and LLM, create vector index, query with similarity search, and print response with sources."]