Poetiq shatters ARC-AGI 2 benchmark at half the cost (poetiq.ai)

Poetiq has achieved a groundbreaking milestone by officially surpassing the ARC-AGI-2 benchmark with a performance rate of 54% at a significantly reduced cost of $30.57 per problem. This achievement marks a new state-of-the-art in AI reasoning as it is the first to break the 50% barrier on this challenging test set. By employing a novel approach called learned test time reasoning, Poetiq's meta-system optimized its solution using the capabilities of the newly released Gemini 3 model. This innovation not only outperformed the previous best score of 45%, which cost over double, but also establishes a new Pareto frontier in cost-effective AI problem-solving. This breakthrough is essential for the AI/ML community as it demonstrates the potential to optimize existing frontier models without the need for extensive fine-tuning. Poetiq's meta-system adapts and improves over time, enhancing its effectiveness for diverse tasks, thereby addressing complex real-world problems that traditional models struggle with. As Poetiq collaborates with partners to solve AI challenges, the implications of their technology could pave the way for more efficient and accessible AI solutions across various industries.