Skip to main content
Local LLM Guide

Choose a local model

The useful model is the one that fits your machine, solves the job, and leaves enough memory for context and normal work. Start with the shortlist rather than the largest download.

Loading model filters and comparison controls…

Read the model guides

Open a guide for its downloads, license, setup choices, and limits. Chat models produce replies; embedding models turn text into searchable representations and do not replace a chat model. Use the filters above to compare memory for your actual computer, compression, and context.

  • Meta Llama 3.1 8B Instruct

    balanced local general chat; multilingual assistant use.

  • Meta Llama 3.2 3B Instruct

    small local assistant deployments; summarization and prompt rewriting.

  • BGE-M3

    Multilingual RAG and semantic search.

  • Nomic Embed Text v1.5

    Small local English embedding model for RAG, similarity search, clustering, and classification.

  • Qwen2.5-Coder 7B Instruct

    Local coding assistant use on 16GB RAM machines; Code generation, code reasoning, and code repair at a small local footprint.

  • Qwen3 8B

    Balanced local Qwen3 reasoning on 16GB RAM systems; Agent and tool-use experiments using the Qwen3 family.

  • Gemma 4 E4B Instruct

    Best default small Gemma 4 model for 16GB laptops.

  • Gemma 4 12B Instruct

    Best for 32GB laptops or desktops, especially Apple Silicon and consumer GPUs.

  • Qwen3.5 4B

    First local multimodal Qwen model on 8GB or 16GB machines.

  • Qwen3.5 9B

    General-purpose local Qwen on 16GB or 32GB systems.

  • GPT-OSS 20B

    Local reasoning, coding experiments, tool calling, and privacy-minded users with 16GB-plus memory and patience today.

  • Qwen3-Coder 30B A3B Instruct

    Developers wanting local repo help, code completion, debugging, and lightweight agent workflows on serious laptops.

  • Qwen3-VL 4B Instruct

    Budget laptops, classroom documents, UI screenshots, basic visual reasoning, and small private agent experiments locally.

  • Qwen3-VL 8B Instruct

    Screenshots, receipts, PDFs, web pages, UI agents, chart reading, and everyday visual reasoning tasks locally.

  • Phi-4 Mini Instruct

    Fast local chat, function-calling tests, multilingual prompts, and long-context experiments on modest consumer hardware today.

  • Qwen3 Embedding 0.6B

    High-quality multilingual document search where long chunks and instruction prompting improve retrieval accuracy locally reliably.

  • DeepSeek-R1-0528-Qwen3-8B

    Local reasoning, math, coding help, and explainable answers on ordinary laptops or small desktops offline.