According to Anthropic (@AnthropicAI), both Claude 4 models have achieved state-of-the-art results on SWE-bench Verified, a benchmark evaluating how AI models resolve real-world software engineering issues (source: AnthropicAI Twitter, May 22, 2025). Claude Opus 4, recognized as the top coding model, demonstrates the ability to work continuously for hours on complex, long-running programming tasks. This advancement in AI agent capabilities is expected to accelerate the adoption of AI-powered tools in blockchain development and smart contract auditing, potentially boosting the value of crypto AI tokens and related projects.