truemodel.tech

Find out which model an endpoint really serves, benchmark it, and write a working config for Claude Code CLI at ~/.claude/settings.json.

How it works — the method, and its limits
Saved API Key Configure
No saved key — enter an endpoint and key below, or save one in Settings

Provider connection

Offline
Will request: https://api.anthropic.com/v1/messages
🔒 Your key is only proxied through this app's server for the request you run — never stored, logged, or sent to any third party. Use “Remember key” below only on a device you trust.

Model testing

How it works

Point it at an endpoint you want to check and it tells you which model is really behind the id — and whether the provider is being honest about it.

  1. 1Paste an endpoint and keyAny OpenAI- or Anthropic-compatible base URL. Your key is proxied server-side and never stored unless you ask it to be.
  2. 2Connect and discoverThe tool lists every model the endpoint actually offers — you don’t need to know the names up front.
  3. 3Test, then identifyBenchmark latency and throughput, then run the identity probe to fingerprint what each id really serves.
  4. 4Read the verdictEach model gets a plain-language verdict: as advertised, wrong tier, or a different model entirely.
Did you know?

A provider can advertise a premium model and quietly route your calls to a cheaper one. The reply still looks plausible, so you rarely notice it by eye.

Enter an endpoint and key in the panel, then Connect & discover to begin.