Challenge the newest AI models with your hardest PhD-level exercises

— and learn how to use AI in your math research —

Sep 1, 2025: We are releasing our first public benchmark Press Release(PDF)

While all models fail the majority of the problems posed by professional researchers, the results reveal wide variation in performance.

View the public benchmark, or browse several sample prompts including their model answers.

Let the best models solve your exercises

Study and solve the exercises of others

Show that your exercises go far beyond the capabilities of LLMs

Challenge the following models with your prompts:

DeepSeek-V3.1 GPT-5 Claude Opus 4.1 Grok-4 Gemini 2.5 Pro o3

Want to benchmark your models against PhD-level math problems from professional researchers?

contact@science-bench.ai