AI-NativeBench: An Open-Source White-Box Agentic Benchmark Suite for AI-Native Systems

TL;DR

This paper addresses the inadequacy of black-box benchmarks for evaluating AI-Native systems, where autonomous, probabilistic agents operate within complex inter-agent protocols. It introduces AI-NativeBench, the first application-centric, white-box benchmark suite grounded in MCP and A2A, employing a trace-first methodology via OpenTelemetry to analyze both performance and reliability across eight applications and 21 system variants. Key findings include a parameter paradox where lightweight models often adhere better to protocols than flagship models, pervasive inference dominance that makes LLM compute the main latency driver, and an expensive failure pattern driven by self-healing retries that inflate token costs. The work provides evidence-based guidance for architectural design, performance engineering, and budget-aware governance, and it releases the benchmark and dataset as open resources to accelerate reproducibility and further research in engineering reliable AI-Native systems.

Abstract

The transition from Cloud-Native to AI-Native architectures is fundamentally reshaping software engineering, replacing deterministic microservices with probabilistic agentic services. However, this shift renders traditional black-box evaluation paradigms insufficient: existing benchmarks measure raw model capabilities while remaining blind to system-level execution dynamics. To bridge this gap, we introduce AI-NativeBench, the first application-centric and white-box AI-Native benchmark suite grounded in Model Context Protocol (MCP) and Agent-to-Agent (A2A) standards. By treating agentic spans as first-class citizens within distributed traces, our methodology enables granular analysis of engineering characteristics beyond simple capabilities. Leveraging this benchmark across 21 system variants, we uncover critical engineering realities invisible to traditional metrics: a parameter paradox where lightweight models often surpass flagships in protocol adherence, a pervasive inference dominance that renders protocol overhead secondary, and an expensive failure pattern where self-healing mechanisms paradoxically act as cost multipliers on unviable workflows. This work provides the first systematic evidence to guide the transition from measuring model capability to engineering reliable AI-Native systems. To facilitate reproducibility and further research, we have open-sourced the benchmark and dataset.

Figures (30)

• Figure 1: Comparison of Cloud-Native and AI-Native application architecture.
• Figure 2: Comparison betweem agent trajectory and trace view.
• Figure 3: Email Responder(MCP).
• Figure 4: Markdown Validator(MCP).
• Figure 5: Book Writer(H-A2A).