A new, challenging AGI test stumps most AI models

A new AI test is outwitting OpenAI, Google models, among others

AI models from OpenAI, Google, DeepSeek, and others scored pathetically low on the second edition of the ARC-AGI benchmark, suggesting models still have a long way to go before achieving AGI.

A new, challenging AGI test stumps most AI models

The Arc Prize Foundation has a new test for AGI that leading AI models from Anthropic, Google, and DeepSeek score poorly on.

The Arc Prize Foundation (co-founded by the famous IA researcher François Chollet) has just launched a new formidable challenge to evaluate general artificial intelligence. Named ARC-AGI-2, this test composed of visual puzzles puts to great test the adaptability and reasoning of the current IA! The results are without appeal: even the ... Read more The article Dearest AGI, but where do you stand?: even the most advanced IA models fail in the fac…

The Arc Prize Foundation says its new ARC-AGI-2 test stumps most AI models; humans get 60% of the questions right but GPT-4.5 and Claude 3.7 Sonnet score ~1%

Maxwell Zeff / TechCrunch: The Arc Prize Foundation says its new ARC-AGI-2 test stumps most AI models; humans get 60% of the questions right but GPT-4.5 and Claude 3.7 Sonnet score ~1%  —  The Arc Prize Foundation, a nonprofit co-founded by prominent AI researcher François Chollet, announced in a blog post …

ARC Prize launches its toughest AI benchmark yet: ARC-AGI-2

ARC Prize has launched the hardcore ARC-AGI-2 benchmark, accompanied by the announcement of their 2025 competition with $1 million in prizes. As AI progresses from performing narrow tasks to demonstrating general, adaptive intelligence, the ARC-AGI-2 challenges aim to uncover capability gaps and actively guide innovation. “Good AGI benchmarks act as useful progress indicators. Source

The ARC Prize Foundation has just released a new, revised and corrected version of its ARC-AGI General Artificial Intelligence Test. This milling remains relatively simple for humans, but chatbots get particularly bad scores.