AgentScout / Best lists / Best Local AI Models & Runtimes

Best Local AI Models & Runtimes (2026)

Open models and local runtimes for running LLMs on your own computer or server, often as the engine behind self-hosted agents. Below are all 9 matching tools in the AgentScout directory, ranked.

How we ranked: Ranked by GitHub stars as recorded on each AgentScout listing (link checks 2026-10-18). Tools without a tracked public repo come after, in alphabetical order. Stars measure developer popularity, not quality or fit for your job, so read each listing before you choose. No paid placements.

  1. Ollama logo

    1. Ollama

    models & local runtimesOpen source★ 183,684 GitHub stars

    Runs open models such as Qwen, Gemma, DeepSeek and gpt-oss locally or in Ollama's cloud, and connects them to agents like Claude Code, OpenClaw and Codex.

    Pricing (as stated): Open source · Mature

  2. LocalAI logo

    2. LocalAI

    models & local runtimesOpen source★ 49,586 GitHub stars

    Self-hosted engine with an OpenAI-compatible API for running LLM, vision, voice and image models on your own hardware, including agents.

    Pricing (as stated): Open source · Mature

  3. Jan logo

    3. Jan

    models & local runtimesOpen source★ 44,988 GitHub stars

    Open-source ChatGPT alternative that downloads and runs LLMs fully offline on your computer, with optional connections to cloud models.

    Pricing (as stated): Open source · Mature

  4. Oumi logo

    4. Oumi

    models & local runtimesOpen source★ 9,400 GitHub stars

    Open platform for fine-tuning, evaluating and deploying open agentic LLMs and VLMs.

    Pricing (as stated): Open source · Mature

  5. MiroThinker logo

    5. MiroThinker

    models & local runtimesOpen weights★ 8,420 GitHub stars

    Deep research agent and open models tuned for complex research and forecasting tasks.

    Pricing (as stated): Open weights · Mature

  6. GLM-5 logo

    6. GLM-5

    models & local runtimesOpen weights★ 7,350 GitHub stars

    Z.ai's open-weights GLM-5 model family built for coding and long-horizon agentic tasks, also offered through the Z.ai API.

    Pricing (as stated): Open weights · Mature

  7. ASI:One logo

    7. ASI:One

    Agentic LLM and chat app from Fetch.ai built to reason, plan and call other agents on the ASI Alliance network.

    Pricing (as stated): Freemium · Active

  8. LM Studio logo

    8. LM Studio

    Desktop app for running local and open models, with Bionic, its agent for work and code that creates documents, slides, PDFs and software.

    Pricing (as stated): See site · Active

  9. Meta Muse Wonder logo

    9. Meta Muse Wonder

    Meta's new models for personal and professional agent tasks, available via Meta Model API and Muse Code.

    Pricing (as stated): Paid API / subscription · Released Oct 12

Browse by category

Related: Running agents locally · Local AI agents · Self-hosted AI agents

More best-of lists: All best lists · Best AI Coding Agents · Best AI Agent Frameworks · Best AI Browser Agents · Best Open-Source MCP Servers · Most-Starred Open-Source AI Agent Projects · Best Self-Hosted AI Agents · Best Free AI Agents (Free Plans) · Best No-Code AI Agent Builders · Best AI Voice Agent Platforms · Best AI Vision Tools for Agents · Best AI Agent Observability & Eval Tools · Best Personal AI Agents · Best AI Agents for Business Teams · Best Crypto AI Agents · Best AI Productivity Agents · Best AI Agent Memory & RAG Tools · Best AI Agent Security & Governance Tools · Best AI Sales & Marketing Agents · Best AI Customer Support Agents

Facts come from our directory listings; pricing is as each maker states it and GitHub stars are as recorded on each listing. Check official sites before you choose.

Get the best new AI agents, weekly

One short email with new tools, launches and what's worth trying. Free, unsubscribe anytime.