Claude Model Detect
深度检测你的 Claude API 真实性。验证模型身份、Thinking 签名、SSE 流完整性。
Your API key is proxied through our Cloudflare Worker only for forwarding. It is never stored, logged, or cached.
About Claude Model Detection
Claude Model Detect is a free tool that verifies whether a Claude API endpoint returns authentic model responses. It performs 17 automated tests across 5 dimensions to detect common proxy fraud tactics.
Why detection matters: Some API resellers cut costs by routing expensive model requests through cheaper models, stripping thinking content, injecting hidden system prompts, or truncating outputs. This tool exposes these practices through protocol-level verification.
How it works: The tool makes 6 carefully crafted API requests and analyzes the responses for Anthropic-specific signatures: msg_ ID prefixes, SSE event sequences, thinking block cryptographic signatures, and token counting accuracy.
FAQ
What does Claude Model Detect check?
It verifies 17 API characteristics across 5 dimensions: model identity, SSE streaming protocol, extended thinking signatures, injection detection, and token consistency. These tests determine if a Claude API proxy returns authentic model responses.
How does it detect fake Claude models?
By checking API response signatures that are hard to forge: the msg_ ID prefix format, SSE event sequence compliance, thinking block content and cryptographic signatures, and token counting accuracy. A proxy using a different model will fail multiple tests.
Is my API key safe?
Yes. Your API key is forwarded through our Cloudflare Worker proxy only for the duration of each test request. It is never stored, logged, or cached. The proxy exists solely to avoid browser CORS restrictions.
What Claude models are supported?
All Claude models available through the Anthropic Messages API, including Claude Sonnet 4, Claude Opus 4, Claude Haiku, and their date-stamped variants. Extended thinking tests require models that support the thinking parameter.
Why would a proxy return a fake model?
Some API resellers cut costs by routing expensive model requests (e.g., Opus) through cheaper models (e.g., Sonnet or Haiku), or by using non-Claude models entirely. They may also strip thinking content, inject system prompts, or truncate outputs.
How many API credits does a test cost?
A full test makes 6 API requests with modest token usage: approximately 2,000-3,000 total tokens. At standard Claude pricing, this costs less than $0.01 per test run.