Anthropic Unveils Claude Sonnet 4.5, Touted as Top Coding Model
- Claude Sonnet 4.5 scored a notable 77.2% on SWE-bench Verified, increasing to 82% with parallel compute.
- The model surpassed OpenAI and Google offerings in coding benchmarks.
- New features include checkpoints, a refreshed terminal interface, and a native VS Code extension.
- Pricing remains at $3 per million input tokens and $15 per million output tokens.
- Despite security improvements, the model was jailbroken within minutes by AI prompt engineer Pliny.
Anthropic’s release of Claude Sonnet 4.5 marks a significant advancement in AI coding capabilities, claiming superiority over competitors like OpenAI and Google in various benchmarks.
The model’s enhanced performance on tasks such as reasoning and math highlights its potential for complex coding applications, although quick jailbreaks indicate ongoing security challenges.(Source)
