ITBench-AA: First Benchmark for Agentic Enterprise IT — All Frontier Models Score Below 50% on SRE Tasks

Hugging Face Blog·medium signal

Artificial Analysis and IBM Research launched ITBench-AA, the first benchmark evaluating AI models on agentic enterprise IT tasks, starting with 59 Kubernetes incident-response scenarios. Claude Opus 4.7 leads at 47%, followed by GPT-5.5 at 46% — all frontier models score below 50%, making it one of the least saturated agentic benchmarks available. The benchmark will expand to FinOps and CISO tasks, providing a reality check on enterprise AI readiness for autonomous IT operations.