Anthropic has released Claude 3.7 Sonnet, which features hybrid reasoning, enhanced coding capabilities, and the ability to output 128K tokens. This new model reportedly surpasses OpenAI's GPT-4o in coding tasks and demonstrates advanced skills in playing Pokémon Red. In a related development, researchers at Hao AI Lab have begun benchmarking various AI models, including Claude 3.7, Claude 3.5, Gemini 1.5 Pro, and GPT-4o, using the Super Mario Bros. game. The tests aim to evaluate the models' performance in real-time scenarios, revealing notable differences in their reasoning and speed. The benchmarking initiative has sparked interest in how different AI systems handle the challenges presented by classic video games.