About local-llm.net

local-llm.net helps people decide whether, and how, to run a real AI workload under their own constraints. We publish free, independent guides to the whole local AI ecosystem, not just our own tools, and we offer paid assessment, deployment and support for teams that need it done properly.

Our Mission

Useful AI shouldn't require sending your data to a third party. Run well, local AI gives individuals, developers and organisations more control: data processed on hardware they choose, no per-token fees, and the ability to work offline. Run badly, it is slower, costlier and less private than people assume. We try to make the difference measurable.

We built local-llm.net to be the definitive resource for anyone deploying AI locally. Whether you're running your first chatbot on a laptop or scaling an enterprise deployment, we have guides, comparisons, and tools to help.

What We Cover

• 50 tools across inference engines, desktop apps, web UIs, developer frameworks, and more
• Step-by-step guides for every platform (Windows, Mac, Linux, Docker, mobile)
• Comparisons with evidence labelled as measured, vendor-reported or estimated (how we label evidence)
• Developer resources including tutorials, code examples, and SDK documentation
• Community contributions including benchmarks, reviews, and experience reports

Built by Cognisoc

The Cognisoc team builds open-source tools for the local AI ecosystem:

• Mullama — Multi-language local LLM inference engine
• Llamafu — Flutter plugin for on-device AI on Android and iOS
• ZigLLM — Educational transformer implementation in Zig

All projects are MIT licensed and available on GitHub.

Neul Labs Limited

local-llm.net is owned and operated by Neul Labs Limited, an Edinburgh engineering company that helps teams put AI into production. Registered in Scotland, company no. SC767862.

Working with us

We sell a working outcome under your constraints. Local inference is one way to deliver it, and sometimes the right recommendation is to stay hosted. Engagements start with a fixed-fee fit assessment and can continue to a pilot, production deployment and managed support. See services and solutions, or tell us about your workload.

How we keep commercial interests away from technical conclusions: editorial & commercial policy.

Get Involved

local-llm.net is community-driven. Contribute guides, submit benchmarks, review tools, or join the conversation on Discord.

Editorial standards

Every model page, tool page, and comparison on this site is meant to stand on its own without a reader needing to double-check it elsewhere. That means we link to primary sources — a model's own model card, a project's own README or license file — rather than restating a spec we can't point back to, and where we don't have a verified number (a benchmark score, a VRAM figure, a licence term) we say what we know instead of guessing at what we don't.

We also try to be explicit about what a recommendation is actually based on. Our "best of" guides name the use case a pick was chosen for rather than presenting one model or tool as universally superior, because almost nothing in local AI is — a model that wins on coding benchmarks can lose badly on creative writing, and a GPU that's the best value for a 24GB build is the wrong recommendation entirely for someone with an 8GB card. Our comparison pages aim to say plainly where a competing tool wins, not just where our own projects do.

How we handle our own projects

Mullama, Llamafu, and ZigLLM are Cognisoc's own open-source projects, and they appear on this site alongside every other tool in their category rather than in a separate promotional section. Where one of them is genuinely the strongest option for a use case — Llamafu for Flutter-based mobile inference, for instance — we say so; where a third-party tool like Ollama or llama.cpp is the better fit, our comparison and "best of" pages say that instead. The goal is a site useful enough that a reader with no interest in Cognisoc's own tools still has reason to come back.