AI Model Check
Check whether an endpoint delivers the model and capabilities it claims. Test any OpenAI or Anthropic-compatible provider with your own API key.
- Your key is used for this run only
- Reopen or share by report code
Every check covers
Compatibility
Response format, token usage, errors and streaming
Capabilities
Tool calls, vision, JSON output and other published features
Model identity
Tokenizer and answers compared with the official API
—
Endpoints observed
—
Completed checks
—
Runs in the last 30 days
—
Potential mismatches
Recent AI model checks
Only completed runs whose owners opted in appear here.
Loading recent activity…
Evidence before you trust an endpoint
A model can be renamed without changing the underlying engine. We combine protocol checks, capability tests and comparisons against baseline behaviour to surface inconsistent claims.
Compatibility
Validate response shapes, token accounting, errors and terminal streaming events.
Capabilities
Exercise real tool calls, vision, JSON output and other published features.
Model identity
Compare tokenizer patterns and deterministic answers against a baseline captured from the official API.
These are behavioural signals, not cryptographic proof of model identity. Gateways, model updates and connection errors can affect results. Review individual checks before drawing conclusions.
How to run a check and read the report