Awesome Free LLM APIs IR — verification evidence policy

A Provider classification may distinguish these layers:

1. Reachability: DNS/TLS/HTTP path is reachable.
2. Signup: an account can be created through the official flow.
3. Key issuance: the official dashboard issues an authorized API key.
4. Inference: an authorized or intentionally anonymous model request returns a valid model response.

HTTP 401, 403, 404, TLS success, website access, timeout or connection refusal alone do not prove successful or blocked model inference.

Direct Iran, authorized VPN/supported-route and credential results are recorded as separate evidence classes. Unknown remains unknown until evidence is conclusive.

Publication safety:
- never publish API keys, Authorization headers, account identities, complete test-host IP addresses or private infrastructure details;
- publish only the minimum evidence required to explain the classification;
- include the check date, request layer and evidence limitation.

Operational use:
- deploy and UX smoke checks are read-only and do not upgrade Provider evidence;
- a healthy website deployment does not prove a Provider account, key or inference path works;
- research sessions must test the deployed revision rather than an unverified local or stale build.

Canonical project: https://llm.persiantoolbox.ir/
Source policy: https://github.com/alirezasafaei-dev/awesome-free-llm-apis-ir/blob/main/docs/IRAN_VERIFICATION_2026-07-20.md