Topic: ai model limitations

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

Salesforce AI Research has introduced MCP-Universe, an open-source benchmark that evaluates large language models' performance in real-world enterprise scenarios, focusing on tool integration and multi-step reasoning. Initial testing revealed that even top models like OpenAI's GPT-5 struggle sign...