Anthropic has released Claude Sonnet 4.5, calling it their strongest coding model yet. It targets developers, researchers, and enterprises needing long-horizon reasoning, code execution, and complex agent building. Available worldwide today through the Claude API at the same $3/$15 per million tokens pricing as Sonnet 4, the model leads benchmarks like SWE-bench Verified and OSWorld, sustaining focus for 30+ hours on multi-step coding tasks.
The rollout includes major upgrades:
- Checkpoints in Claude Code
- A VS Code extension
- Context editing and memory tools in the API
- File creation in the Claude apps
- The Claude for Chrome extension for Max users
- Imagine with Claude
BREAKING 🚨: Anthropic released "Imagine with Claude", where Claude can generate software on the fly!
— TestingCatalog News 🗞 (@testingcatalog) September 29, 2025
This is the future of the internet 👀
Sadly, "Available to Max users for 5 days" https://t.co/bPZaLFcnUI pic.twitter.com/WgF5lNdbLc
Anthropic made its Claude browser extension available to MAX users who signed up for the waitlist last month.
— TestingCatalog News 🗞 (@testingcatalog) September 29, 2025
MAX FOMO 👀 pic.twitter.com/O2210jzmHG
Anthropic also unveiled the Claude Agent SDK, offering the same infrastructure used internally.
Early adopters from GitHub Copilot, Canva, Figma, Devin, and security teams report large productivity and accuracy gains, from debugging to financial analysis and red teaming. The model also incorporates expanded safety training and defenses against prompt injection, making it the most aligned Claude release to date.
