Anthropic/Model/Claude/Sonnet/4.5
Key Features
- Released on September 29, 2025
- Described as "best coding model in the world"
- Strongest model for building complex agents
- Superior performance in computer use tasks
- Significant improvements in reasoning and math
- Enhanced domain-specific knowledge in finance, law, medicine, and STEM
- Knowledge cutoff: September 2025
- Model version status: Stable release
Performance Highlights
- SWE-bench Verified: 77.2% (averaged over 10 trials)
- OSWorld benchmark: 61.4% (up from 42.2% four months prior)
- Can maintain focus for over 30 hours on complex tasks
- Reduced misaligned behaviors like sycophancy and deception
- Most aligned frontier model according to Anthropic
Model Tiers
- claude-sonnet-4-5
- Full-featured flagship model
- Best performance across all metrics
- Previous versions:
Access Details
- API availability: Available via
claude-sonnet-4-5 in the Claude API - Pricing:
Inline: 3/15 per million tokens (same as Claude Sonnet 4) - Released under AI Safety Level 3 (ASL-3) protections
- Available across all Claude products (API, web, apps)
Technical Specifications
- Model architecture: Details not specified
- Training data: Details not specified
- Input/output formats: Text, images, code
- Special capabilities:
- Enhanced coding abilities
- Improved reasoning
- Complex agent building
- Long context focus (30+ hours)
- Native VS Code extension
- Context editing and memory tools
- Code execution and file creation in Claude apps
References