AI is specialized, and the model every vendor calls “the best” was not measured on your job. XF Model Source works out what your job really requires, judges 450+ OpenRouter models on more than 200 ways they differ, then tests the best few on your own tasks with your own OpenRouter key. Discovery and the shortlist are free. XF Model Watch, the paid add-on, re-tests your saved jobs every week.
Part of Model Source
Connect with GitHub — it's freeThe full product lives at xfms.xpansion.dev. Once XFMS is connected to your AI assistant you just ask — “help me pick an AI model for [your job]” — and it asks a few plain questions, works out what the job really needs, and comes back with a shortlist and the evidence behind it.
The server is hosted, so there is nothing to install. The address is https://xfms.xpansion.dev/mcp
Claude Code — paste this in your terminal, then type /mcp inside Claude Code to sign in:
claude mcp add model-source --transport http https://xfms.xpansion.dev/mcp
Codex (in the ChatGPT desktop app) — Settings, MCP servers, Add server. Name it model-source, choose Streamable HTTP, paste the address, save, then select Restart. When it shows Authenticate, select it and sign in with your email.
Codex (in a terminal) — paste these two lines (the second signs you in):
codex mcp add model-source --url https://xfms.xpansion.dev/mcp
codex mcp login model-source
Claude (claude.ai or the desktop app) — Settings, Connectors, Add custom connector, paste the address.
Cursor, Windsurf and other assistants — add an HTTP MCP server at the address (Cursor: ~/.cursor/mcp.json):
{
"mcpServers": {
"model-source": { "url": "https://xfms.xpansion.dev/mcp" }
}
}
Your assistant opens a sign-in page the first time. If yours can’t, sign in at xfms.xpansion.dev/signin and use the personal connection link on your account page.
What it costs. Discovery and the shortlist are free. Tests on your own tasks run on your OpenRouter key, usually a few cents; you see the most a test can cost before anything runs. XF Model Watch (optional, from $19 a month) re-tests your saved jobs every week and emails you only when a cheaper or better model passes every one of your checks.
You describe the job. XFMS asks a few plain questions and turns your answers into the job's real requirements: what goes in, what must come back, the format, the cost ceiling, what must never happen. You don't need to know what to ask for; neither does a leaderboard.
Independent test scores, each model's published capabilities and price, and the results of our own comparisons. For every requirement a model is measured, judged from related tests, or shown as unknown. Unknown is never shown as good.
A screening round runs about 12 likely models on one of your sample tasks. Any model that breaks a must-have is ruled out, and the best six run all your tasks. You get the winner, the reasons, and every answer so you can check it yourself. Tests run on your own OpenRouter key.
Models, prices and retirements change every month. Watch re-tests your saved jobs weekly and emails you only when a cheaper or better model passes every one of your checks, with a warning before your model is retired.
You describe your job in plain words. Concrete is better: 'pull invoice fields into JSON' works; 'extraction' does not.
XFMS asks only the questions that would change the answer, and writes down what the job really requires.
It checks 450+ OpenRouter models against those requirements, using independent test scores, published capabilities, prices and our own comparison results.
You get a shortlist with the evidence behind each pick. Anything without evidence is shown as unknown.
A screening round tests about 12 likely models on one of your tasks; the best six run all of them on your own OpenRouter key. You see the maximum cost first, and nothing runs until you say go.
Optional: XF Model Watch repeats the test every week and tells you when a better or cheaper model passes.