AI BriefingClaude 4
Key point
Anthropic has unveiled Claude 4, its next-generation model family with dramatically enhanced coding and reasoning capabilities.
Details
Anthropic announced its next-generation models Claude Opus 4 and Claude Sonnet 4, which maximize coding, advanced reasoning, and AI agent performance.
Claude Opus 4 is the world's best coding model, achieving leading scores on SWE-bench (72.5%) and Terminal-bench (43.2%). It can sustain focus on complex, long-running tasks spanning thousands of steps for hours at a time without performance degradation.
Claude Sonnet 4 records 72.7% on SWE-bench, offering an optimal balance of performance and efficiency. GitHub plans to adopt it as the core model for its new coding agent.
Key new features and updates include the following:
- Extended thinking with tool use (beta): Interleaves tools such as web search during the reasoning process to improve response accuracy.
- Parallel tool use and improved memory: Uses multiple tools simultaneously, and when developers grant local file access, extracts and stores key information to maintain continuity.
- Claude Code general availability (GA): Integrates with GitHub Actions, VS Code, and JetBrains for seamless pair programming.
- New API capabilities: Provides a code execution tool, the MCP connector, the Files API, and prompt caching of up to one hour.
Both models are available via Amazon Bedrock and Google Cloud Vertex AI. Pricing remains the same as previous models: Opus 4 at $15/input and $75/output (per 1M tokens), and Sonnet 4 at $3/input and $15/output.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.