ARC-AGI-3 offers $2M to any AI that matches untrained humans, yet every frontier model scores below 1%
AI Summary
The ARC-AGI-3 benchmark, as reported by The Decoder, has introduced a $2 million prize for any AI system that can match the performance of untrained humans on its tasks. The benchmark places AI systems into interactive game environments that humans solve with relative ease, yet every current frontier AI model scores below 1% on the evaluation. The benchmark is specifically designed to neutralize the typical advantages of large AI models, making it a particularly stringent test of general reasoning and adaptability. The sub-1% performance across all frontier models underscores a significant gap between current AI capabilities and human-level general intelligence, despite recent advances in the field.
Why it matters
The ARC-AGI benchmark series, developed by François Chollet and the ARC Prize Foundation, has become a closely watched measure of progress toward artificial general intelligence (AGI), making each new iteration a key reference point for assessing the true capabilities of leading AI systems from companies such as OpenAI, Google, and Anthropic. The near-zero scores across all frontier models on ARC-AGI-3 challenge prevailing narratives about rapid AGI progress and may influence investor and analyst sentiment regarding the timeline and feasibility of transformative AI breakthroughs. The $2 million prize structure also signals continued private-sector and research-community investment in defining and incentivizing measurable AGI milestones, a dynamic with long-term implications for AI research funding and competitive positioning.
Scoring rationale
A major new AGI benchmark with a $2M prize directly measures frontier AI model capabilities, with broad implications for the AI industry and companies competing in the space, though it lacks immediate direct market impact.
Impacted tickers
This summary was generated by AI from the original article published by The Decoder. AIMarketWire does not provide trading advice. Always refer to the original source for complete reporting.