Artificial intelligence systems may be good at generating text, recognizing images, and even solving basic math problems—but when it comes to advanced mathematical reasoning, they are hitting a wall.

First Proof benchmark results released October 14, 2026, show frontier AI systems solved 90% of unpublished research math ...

ORCA benchmark trips up ChatGPT-5, Gemini 2.5 Flash, Claude Sonnet 4.5, Grok 4, and DeepSeek V3.2 In the world of George Orwell's 1984, two and two make five. And large language models are not much ...

The FrontierMath benchmark from Epoch AI tests generative models on difficult math problems. Find out how OpenAI’s o3 and other AI models performed. FrontierMath accuracy for OpenAI’s o3 and o4-mini ...

OpenAI publishes 722 AI-generated math papers after its models outgrew benchmarks. See what's solved and what's unverified.

AI benchmarks are useful in assessing AI model performance. But when most developers report high scores, benchmarks become less meaningful. A recent study found that some large AI companies privately ...

Researchers are racing to develop more challenging, interpretable, and fair assessments of AI models that reflect real-world use cases. The stakes are high. Benchmarks are often reduced to leaderboard ...

Are AI benchmarks really the gold standard we’ve been led to believe? Matt Wolfe walks through how these widely accepted metrics, designed to measure the performance of artificial intelligence systems ...