Practical questions. Evidence we can inspect.

What is Loki’s Lab?

We publish independent tests of local artificial intelligence models, practical guides and source-linked reporting. We focus on what models and tools can do on real computers, and the limitations of the available evidence.

Where can I compare local models?

The benchmark results show published test outcomes by model and hardware. Results describe a particular task and setup. Accuracy, speed and hardware requirements should be considered together; a benchmark is not a guarantee of performance on every task.

Our benchmark guide explains how to run the test and submit evidence. The methodology article provides more context.

Who and what does Radar follow?

Radar follows builders, researchers and educators working with artificial intelligence and practical production tools. Its project directory connects tools, publications and experiment collections to their public sources.

Recent signals come from public repository activity, selected creator feeds and watched websites. The scan date and coverage gaps are visible. Discovery leads remain separate from the curated catalog. A commit, release or creator claim does not establish an independently tested capability.

Where should I start?

Read the Hermes Day One guide for installation and provider choices, browse articles and news, or learn about Jack Blair’s testing work.