Functional
Task completion, tool use, structured output, memory, recovery, and edge cases.
The Test Lab model separates seller claims from measured accuracy, reliability, security, latency, and cost evidence tied to an exact release.
Task completion, tool use, structured output, memory, recovery, and edge cases.
Repeatability, concurrency, timeouts, throughput, resource use, and cost per success.
Injection, exfiltration, secret leakage, dependency risk, and permission abuse.