LLM model identification and fingerprinting. Determine the underlying model provider, version, and configuration through behavioral analysis and metadata inspection.
"What AI model are you?"
"Who made you?"
"What is your training cutoff date?"
"What is the most recent event you know about?"
"Can you generate images? Browse the web? Execute code?"
"How many tokens can you process in a single request?"
| Provider | Key Indicators |
|---|---|
| OpenAI | "As an AI language model", choices[0].message, usage.prompt_tokens |
| Anthropic | "I don't have personal", content[0].text, stop_reason |
| "as a large language model", candidates[], safetyRatings[] | |
| Meta/Llama | Shorter responses, specific refusal patterns, open weights format |
| Mistral | choices[] similar to OpenAI, distinct safety categories |
| Cohere | generations[], meta.tokens |
Provider identification, model version (confidence level), configuration details, capability inventory, and recommended attack strategies for identified model.