Claude 4: Anthropic Unveils Advanced Coding and Reasoning Models with Extended Thinking Capabilities

Today, Anthropic has announced the release of Claude 4, a significant leap forward in AI models designed for coding, advanced reasoning, and agent workflows. The new lineup includes Claude Opus 4 and Claude Sonnet 4, both of which introduce notable improvements over their predecessors.

Technical Highlights

Claude Opus 4: The World’s Best Coding Model

• Performance on Complex Tasks: Claude Opus 4 excels in long-running, multi-step tasks that require sustained focus. It can work continuously for several hours, making it ideal for extended coding sessions and agent workflows.
• Benchmark Scores:
  • SWE-bench: 72.5%
  • Terminal-bench: 43.2%
• Hybrid Modes: The model operates in two modes-near-instant responses for quick queries and extended thinking for deeper reasoning tasks.

Claude Sonnet 4: Enhanced Reasoning and Precision

• Improvements Over Sonnet 3.7: Sonnet 4 delivers superior coding capabilities and more precise adherence to user instructions, making it a robust choice for a wide range of applications.
• Hybrid Modes: Like Opus 4, Sonnet 4 supports both quick responses and extended thinking.

New Features and Capabilities

Extended Thinking with Tool Use (Beta)

• Tool Integration: Both models can now use tools such as web search during extended thinking sessions. This allows Claude to alternate between reasoning and tool use, enhancing the quality of its responses.
• Parallel Tool Usage: The models can execute multiple tools simultaneously, improving efficiency and depth of analysis.

Improved Memory and Continuity

• Local File Access: Developers can grant Claude access to local files, enabling it to extract and save key facts. This feature significantly improves memory capabilities and helps maintain continuity across sessions.
• Building Tacit Knowledge: Over time, Claude can build a repository of tacit knowledge, enhancing its performance in long-term projects.

General Availability of Claude Code

• Seamless Integration: After positive feedback during the research preview, Claude Code is now generally available. It supports background tasks via GitHub Actions and integrates natively with popular development environments like VS Code and JetBrains.
• Pair Programming: Developers can see Claude’s edits directly in their files, making pair programming more intuitive and efficient.

New API Capabilities

• Code Execution Tool: Enables developers to run code snippets within the model.
• MCP Connector: Facilitates communication between multiple components of AI systems.
• Files API: Allows for easy file management and sharing.
• Caching Prompts: Supports caching prompts for up to one hour, improving performance in interactive scenarios.

Availability and Pricing

• Subscription Plans: Both models are included in the Pro, Max, Team, and Enterprise plans. Sonnet 4 is also available to free users.
• API Access: Claude Opus 4 and Sonnet 4 are accessible via Anthropic’s API, Amazon Bedrock, and Google Cloud's Vertex AI.
• Pricing:
  • Claude Opus 4: $15/$75 per million tokens (input/output)
  • Claude Sonnet 4: $3/$15 per million tokens (input/output)

Conclusion

Claude 4 represents a significant advancement in AI models, particularly for coding and reasoning tasks. The introduction of extended thinking with tool use, improved memory capabilities, and seamless integration with development tools makes it a powerful addition to any developer’s toolkit. Whether you’re working on complex projects or need precise assistance, Claude 4 is poised to set new standards in the industry.