Skip to content
AI Model Radar

Claude Sonnet 4.5

claude-sonnet-4-5-20250929
Anthropic · reasoning · api · seen 12m ago
GA Alert me on changes
Context
1M
Max output
64K
Input $/1M
$3.00
Output $/1M
$15.00
Modalities
text · image · document
Released
29 Sept 2025
AI summary
● machine-written

Claude Sonnet 4.5: Anthropic's frontier coding and agent model

Claude Sonnet 4.5 is Anthropic's state-of-the-art model for software coding, complex agents, and computer use tasks, released September 29, 2025. It demonstrates significant performance gains on real-world coding benchmarks (SWE-bench Verified) and computer interaction tasks (OSWorld 61.4%), with improved reasoning and domain-specific knowledge across finance, law, medicine, and STEM. The model maintains focus on complex multi-step tasks exceeding 30 hours and supports text, image, and document inputs with a 1M token context window.

What's new
  • State-of-the-art performance on SWE-bench Verified for real-world software coding
  • Computer use benchmark lead at 61.4% on OSWorld (vs. 42.2% for Sonnet 4 four months prior)
  • Enhanced domain knowledge in finance, law, medicine, and STEM with improved reasoning
  • Improved alignment compared to previous Claude models
Best for
Complex multi-step software engineering and debugging across large codebasesAutonomous AI agents requiring sustained coherence and adaptive decision-makingBrowser-based and computer interaction tasksDomain-specific reasoning in finance, law, medicine, and technical fields
Sources

Source: https://docs.claude.com/en/docs/about-claude/models/overview