What did Anthropic do?! (Opus 5)

The video introduces Claude Opus 5, highlighting its impressive performance across various benchmarks, often outperforming competitors like Fable 5 and even GPT-4 on tasks such as coding, problem-solving, and complex enterprise work. The presenter analyzes benchmark results, showing Opus 5's superiority in cost-effectiveness and overall score, particularly on tasks requiring exhaustive reasoning. It also showcases Opus 5's ability to handle tasks like building a wind tunnel simulation and passing the ARC-AGI benchmark. The pricing for Opus 5 is set at $5 per million input tokens and $25 per million output tokens, the same as Opus 4.8. The release also includes updates to the Claude platform and API, such as mid-conversation tool changes and automatic fallbacks. The presenter expresses surprise and excitement about the model's capabilities, especially its strong performance on novel problem-solving tasks at a lower cost.