ARC-AGI-3 Benchmark Shows Top AI Models Fail at General Reasoning, Scoring Below 1%

A new benchmark is forcing a hard question: are today’s most advanced AI models actually capable of reasoning—or just very good at pattern recognition?

Gemini 3.1 Pro Improves AI Reasoning Across Google’s Ecosystem, Targeting Complex Enterprise and Developer Workflows

Google’s newest Gemini model focuses on deeper reasoning — not just faster answers — as AI competition shifts toward intelligence quality.

Google and OpenAI Models Outperform Humans at ICPC Coding Finals

OpenAI’s GPT-5 achieved a perfect score of 12/12, while Google’s Gemini 2.5 Deep Think solved 10/12, ranking second overall at the 2025 ICPC World Finals.

OpenAI’s BrowseComp Tests AI Browsing Skills on Hard-to-Find Questions

Amazon to Launch ‘Reasoning’ AI Model to Rival OpenAI & Anthropic