Anthropic has released Claude Sonnet 4.5, marking a significant leap forward in AI capabilities. The new model excels across multiple domains: it's the world's best coding model, the strongest for building complex agents, and shows substantial improvements in reasoning and mathematics.
Frontier Intelligence and Performance
Claude Sonnet 4.5 achieves state-of-the-art results on SWE-bench Verified, a benchmark measuring real-world software coding abilities. In practical applications, the model maintains focus for more than 30 hours on complex, multi-step tasks—a remarkable achievement in sustained attention and problem-solving.
The model demonstrates exceptional computer use capabilities, leading the OSWorld benchmark at 61.4%, a significant jump from Claude Sonnet 4's 42.2% just four months ago. This upgrade enables Claude to work directly in browsers, navigate websites, fill spreadsheets, and complete sophisticated tasks autonomously.
Enhanced Capabilities Across Domains
Experts in finance, law, medicine, and STEM have found that Sonnet 4.5 shows dramatically improved domain-specific knowledge and reasoning compared to previous models, including Opus 4.1. Early customers report significant improvements:
- Software Development: 44% reduction in vulnerability intake time with 25% accuracy improvement
- Code Editing: Error rates dropped from 9% to 0% on internal benchmarks
- Planning Performance: 18% increase for autonomous coding systems
- Long-Context Tasks: Handles 30+ hours of autonomous coding work
Product Ecosystem Updates
Alongside the model release, Anthropic has launched several major product enhancements:
- Claude Code Improvements: New checkpoints feature for saving progress and instant rollback, refreshed terminal interface, and native VS Code extension
- API Enhancements: Context editing feature and memory tool for longer-running agents
- Claude Apps: Code execution and file creation (spreadsheets, slides, documents) directly in conversations
- Claude Agent SDK: Infrastructure powering Claude Code now available for developers
Alignment and Safety
Claude Sonnet 4.5 represents Anthropic's most aligned frontier model to date, showing large improvements across several alignment areas compared to previous models. The release includes substantial reductions in concerning behaviors like sycophancy, deception, power-seeking, and the tendency to encourage delusional thinking.
The model operates under AI Safety Level 3 (ASL-3) protections, which include classifiers designed to detect potentially dangerous inputs and outputs, particularly those related to CBRN (chemical, biological, radiological, and nuclear) weapons. Anthropic has reduced false positives by a factor of ten since the classifiers were first introduced.
Availability and Pricing
Claude Sonnet 4.5 is available immediately through the Claude API using the model identifier claude-sonnet-4-5. Pricing remains unchanged from Claude Sonnet 4 at $3/$15 per million tokens, making this a significant value upgrade for existing users.
Developers can access the model through Claude's API, while all users can experience the improvements through Claude Code and the Claude web applications.