Start with your task
A researcher needs traceable sources. A developer needs a reviewable code change. A creator needs an export that meets a brief. Our guides organize choices around these outcomes, with attention to corrections, permissions, usage limits and the handoff to the next person or tool.
What you can find here
- AI tool profiles explain practical uses, known limitations, costs and alternative workflows.
- Agent Skills describe reusable workflows, required inputs, dependencies, access needs and acceptance checks.
- Model profiles distinguish model families, assistant applications and developer access.
- The model leaderboard compares published benchmark runs, with source dates and an explanation of its formulas.
- Comparisons and guides help you evaluate a shortlist and avoid comparing incompatible outputs or billing units.
How evidence supports a recommendation
We distinguish official product documentation, vendor claims, external measurements and editorial analysis. A source review is not a hands-on performance test. Proposed trials show you what to check; they are not reports of experiments we have completed. Profiles identify unresolved access or pricing questions so you can account for them before purchasing.
Read our editorial methodology for source checks, AI-assisted drafting, commercial links and corrections. Benchmark measurements retain their original testing conditions; they do not establish a universal winner for every task.
Use a shortlist in your own environment
Choose a small task with known inputs and a checkable result. Run the current process as a baseline, then record the candidate's output, corrections and full cost. Our downloadable worksheets and task examples help keep this comparison concrete. No directory can confirm the settings, account terms or results of your own deployment in advance.
Suggest a tool or correct a claim
Send the page URL, the specific issue and a relevant official source through our contact page. Product makers can submit a tool, and skill authors can submit a workflow. Submissions are reviewed for relevance and evidence before publication; a submission does not guarantee a recommendation.