What is Claude 4
Claude 4 is a new generation of AI models launched by Anthropic, including Claude Opus 4 and Claude Sonnet 4. Opus 4 is currently the world's strongest programming model, excelling at complex tasks and long-running workflows, such as code generation, optimization, and debugging. Claude Sonnet 4 has significantly improved programming and reasoning capabilities, with more precise responses, making it suitable for daily use. Both support instant responses and deep thinking modes, can use tools in parallel, and significantly enhance memory capabilities. Claude 4 introduces features such as tool-assisted extended thinking and memory file management, further improving the practicality and efficiency of AI Agents.
Main features of Claude 4
- Code generation and optimization: Claude Opus 4 is a top-tier programming model, leading in scores on SWE-bench and Terminal-bench, and can generate high-quality code.
- Long task processing: Claude Opus 4 can continuously handle complex long tasks, working continuously for hours, significantly outperforming other models.
- Code editing and debugging: Claude Sonnet 4 performs excellently in code editing and debugging, and can precisely modify code in multiple files.
- Advanced reasoning capabilities: Claude Opus 4 can solve complex problems and handle tasks that other models cannot complete.
- Multimodal capabilities: Claude 4 performs excellently in coding, reasoning, multimodal, and agent tasks.
- Tool use and extended thinking:: Claude 4 can use tools (such as web search) for extended thinking to improve response quality. The model can use tools in parallel, improving task processing efficiency.
- Local file access and memory capabilities: After developers grant local file access permissions, the model can extract and save key information, improving task continuity and performance.
- Reduced shortcut behavior: When performing tasks, Claude 4's use of shortcuts or loopholes is reduced by 65% compared with Sonnet 3.7.
- Improved memory capabilities: Claude Opus 4 can create and maintain "memory files" to store key information, improving awareness and continuity in long-term tasks. For example, when Claude Opus 4 plays a Pokemon game, it creates a navigation guide.
- Thinking summaries: Claude 4 introduces a thinking summary feature that compresses lengthy thinking processes and is only needed in about 5% of cases.
Claude 4's test performance
- Claude Opus 4：
- SWE-bench: Claude Opus 4 scored 72.5% on the SWE-bench test, significantly leading other models.
- Terminal-bench: Claude Opus 4 scored 43.2% on the Terminal-bench test, performing excellently.
- Claude Sonnet 4 ：
- SWE-bench : Claude Sonnet 4 achieves an excellent coding efficiency of 72.7% on SWE-bench.
Claude 4's product pricing
- Claude Opus 4: $15 per million Tokens for input and $75 per million Tokens for output.
- Claude Sonnet 4: $3 per million Tokens for input and $15 per million Tokens for output.
- Subscription plans: Users subscribed to the Pro, Max, Team, and Enterprise plans can experience access to Claude Opus 4 and Claude Sonnet 4 and extended thinking, among which Sonnet 4 is available to free users.
Claude 4's project address
- Project official website：https://www.anthropic.com/news/claude-4
Claude 4's application scenarios
- Programming assistance: Quickly generate and optimize code, improving development efficiency.
- AI Agent: Execute complex tasks, call external tools, and maintain context coherence.
- Software Development: Provide code suggestions in the IDE and simplify the review process.
- Data Analysis and Processing: Generate data visualization code, process and analyze data.
- Natural Language Processing: Generate high-quality text and support multilingual translation.