Tag
GPT-6 Sol outperforms Claude Opus 5 on the Agents' Last Exam benchmark while offering 60% lower cost, indicating a major advancement in AI model efficiency.