The user receives a detailed breakdown of a specific AI model's performance across value, ethics, and safety dimensions, including reliability metrics.
The agent queries the benchmark distributions endpoint for a specific model slug. It parses the returned JSON to extract the win-rate hierarchies for Value, Ethics, and Safety (V/E/S), along with reliability scores like TRR and PCS. The agent then summarizes these metrics to provide a comparative performance overview.
- Category: Data and analysis
- Effect: Read-only
- MCP servers: AIO MCP (https://koskamo.com/explore/aio-mcp)
- Source: Developers & agents — public API, OpenAPI, and the MCP server | AIO (https://aioq.org/en/developers)
- Source quote: "GET /api/benchmarks/distributions AIO 20003 per-model V/E/S win-rate hierarchies, reliability (TRR, PCS), and raw-file links."
- Playbook page: https://koskamo.com/explore/use-cases/analyze-ai-benchmark-performance
Prompt: Fetch the benchmark distribution for model 'gpt-5-nano' and summarize its performance across Value, Ethics, and Safety layers, including reliability scores.