OpenAI closes reasoning gap in voice agents

PLUS: Test multiple AI models with the same prompt, fast