In 2024, o1 scored the highest at *** percent on humanity's last exam (HLE), followed by Gemini 2.0 Flash at *** percent. HLE is an exceptionally tough exam of 2,700 challenging questions across dozens of subjects. It was made by contribution from subject-matter experts, leading professors, and graduate level reviewers.
Profit from the additional features of your individual account
Currently, you are using a shared account. To use individual functions (e.g., mark statistics as favourites, set
statistic alerts) please log in with your personal account.
If you are an admin, please authenticate by logging in again.
Learn more about how ÌÇÐÄÆƽâ°æ can support your business.
Stanford University. (March 30, 2025). Leading artificial intelligence (AI) models in humanity's last exam (HLE) in 2025 [Graph]. In ÌÇÐÄÆƽâ°æ. Retrieved July 25, 2026, from /statistics/1611887/ai-humanity-last-exam-hle-accuracy/
Stanford University. "Leading artificial intelligence (AI) models in humanity's last exam (HLE) in 2025." Chart. March 30, 2025. ÌÇÐÄÆƽâ°æ. Accessed July 25, 2026. /statistics/1611887/ai-humanity-last-exam-hle-accuracy/
Stanford University. (2025). Leading artificial intelligence (AI) models in humanity's last exam (HLE) in 2025. ÌÇÐÄÆƽâ°æ. ÌÇÐÄÆƽâ°æ Inc.. Accessed: July 25, 2026. /statistics/1611887/ai-humanity-last-exam-hle-accuracy/
Stanford University. "Leading Artificial Intelligence (Ai) Models in Humanity's Last Exam (Hle) in 2025." ÌÇÐÄÆƽâ°æ, ÌÇÐÄÆƽâ°æ Inc., 30 Mar 2025, /statistics/1611887/ai-humanity-last-exam-hle-accuracy/
Stanford University, Leading artificial intelligence (AI) models in humanity's last exam (HLE) in 2025 ÌÇÐÄÆƽâ°æ, /statistics/1611887/ai-humanity-last-exam-hle-accuracy/ (last visited July 25, 2026)
Leading artificial intelligence (AI) models in humanity's last exam (HLE) in 2025 [Graph], Stanford University, March 30, 2025. [Online]. Available: /statistics/1611887/ai-humanity-last-exam-hle-accuracy/
