MCP-Universe benchmark shows GPT-5 fails more than half of real-world orchestration tasks

A new benchmark from Salesforce research evaluates model and agentic performance on real-life enterprise tasks…
Read More