AI 發展尚有巨大的空間，沒有又快又便宜，只有更快更便宜。

VERSES AI 公司宣布，其 Genius 模型在名為 Mastermind 的密碼破解挑戰賽中擊敗了中國的 DeepSeek R1。 在破解 100 款遊戲的密碼時，Genius 比 DeepSeek 模型快 245 倍，成本低 779 倍。

VERSES Genius model surpasses China's DeepSeek R1 in code-breaking challenge

VERSES AI Inc. announced that its Genius model outperformed China's DeepSeek R1 in a code-breaking challenge called Mastermind. Genius was 245 times faster and 779 times cheaper than DeepSeek's model in cracking the code across 100 games.

This demonstration highlights Genius's advanced multi-step reasoning capabilities. The challenge required models to deduce the correct code using logical reasoning and feedback, showcasing Genius's efficiency in dynamic problem-solving.

The results suggest that domain-specific models like Genius may offer significant advantages over general-purpose AI models, potentially enhancing the reliability and accuracy of AI agents in real-world applications.