Anthropic recently announced itsNew AI models: Claude Opus 4 and Claude Sonnet 4, boasting of providing multi-step reasoning and program code generation capabilities, while also having instant response capabilities and expanded thinking operation modes.
At the same time, Anthropic also boasts that the two models provide an additional developer mode that can fully display the model's thinking process.
Claude Opus 4 can provide long-term focus in complex workflows to complete multi-step reasoning tasks. It is touted as Anthropic's highest-performing artificial intelligence model and has higher memory capacity, which allows it to cope with longer-term inference operations with contextual content, and can also perform longer-term program rewriting tasks.
Claude Sonnet 4 boasts significant improvements in programming and mathematical calculation performance, can accurately follow user commands, and balances performance output and resource utilization. Its memory and reasoning capabilities have also been enhanced, making it suitable for executing large-scale tasks or serving as an artificial intelligence agent for complex interactions.
In terms of price per million words per input and output, Claude Opus 4 costs $15 for input and $75 for output, while Claude Sonnet 4 costs $3 for input and $15 for output. Claude Sonnet 4 is currently available to all users, while Claude Opus 4 is only available to paid subscribers.
On the other hand, Anthropic also boasts that it has added strict protection to Claude Opus 4 to prevent the artificial intelligence model from giving adverse responses and to prevent users from circumventing relevant regulations through instructions, such as requiring the artificial intelligence to provide specific program code cracking methods, or methods for making and obtaining illegal items.
Anthropic also boasts that Claude Opus 4's security performance has reached the level needed for individuals with STEM backgrounds to acquire biological, chemical, and nuclear weapons, and that the model has adopted the "ASL-3" security level standard. Other upgrades include a new version of the Claude Code tool, supporting extensions for Microsoft VS Code, JetBrains, and GitHub, and allowing developers to create various AI agent services through a new API.
