EGI-BENCH: An Enterprise General Intelligence Benchmark Suite

Enterprise adoption of LLM agents is gated by evaluations that span a wide arc of abilities: CRM operations, heterogeneous-document retrieval, personal data understanding, multi-turn planning, tool use, function calling, long-horizon research, and long-context software engineering. These evaluations live today in separate repositories with divergent runners, dependencies, and scoring con- ventions, which makes cross-benchmark compar- ison difficult and obscures what a single aggre- gate number actually represents. We release EGI- BENCH, a unified benchmark suite that consol- idates eight enterprise-facing agent benchmarks (CRMARENA, HERB, PERSONABENCH, USER- BENCH, MCP-UNIVERSE, MCPEVAL, BENCH- FORCE, LOCOBENCH) behind a single command- line interface with two runtime-budgeted exe- cution modes. On top of the suite we de- fine the Enterprise General Intelligence (EGI) Score, a six-dimensional framework (M1–M6) whose buckets are populated by fine-grained sub- metrics routed from each benchmark rather than by benchmark-level averages. We describe the suite, the sub-metric routing, and the quick/deep budgets, and report results for fourteen fron- tier and cost-efficient LLM agents spanning the GPT, Claude, and Gemini families under quick mode. Top-line EGI ranges from 39.4 (GPT- 4o mini) to 66.6 (gpt-5.5), and no single model dominates every capability axis: gpt-5.5 leads task outcome and planning, Claude Son- net 4.5 leads tool actioning and retrieval, Gem- ini 3.1 Pro Preview leads safety and abstention, and Claude Opus 4.5 leads engineering quality.