SerpApi hits 97.6% accuracy on SimpleQA Verified

SerpApi connected its Google Search and Extract APIs to GPT-6.1 Sol at low reasoning effort and reached 97.6% accuracy on Google DeepMind's SimpleQA Verified benchmark. After excluding 40 questions with leaked answers, the model answered 937 of 960 correctly. The score matches Tavily's published 97.6% and leads Parallel (97%), Perplexity (95%) and Exa (91%).

• Model answered 937 of 960 questions correctly after exclusions
• 40 questions dropped because search results exposed benchmark answers
• Tavily Search also scored 97.6%, Parallel Advanced 97.0%
• On SealQA the setup scored 58.56% (65 of 111)