Claude-4 Sonnet

Entity category: technology

Artificial Intelligence

GPT-5 Fails Over 50% of Real-World Orchestration Tasks in MCP-Universe Benchmark

August 23, 2025
Retro robots sit at desks in a classroom, taking a test. A charming illustration.

Salesforce AI Research has introduced MCP-Universe, an open-source benchmark that evaluates large language models' performance in real-world enterprise scenarios, focusing…

Read More »