Anthropic · 2025 · 02 · 24 · Model · ~1 min read

Anthropic launched Claude 3.7 Sonnet + Claude Code

Anthropic upgraded its mid-tier model with an extended thinking mode you can turn on for harder questions, and shipped Claude Code — a command-line tool that lets developers run Claude as a coding agent in their terminal.

What's actually new

  • Extended thinking toggle. You decide when Claude pauses to think harder. Cheaper than always-on reasoning.
  • Claude Code — a terminal tool. Tell it to fix a bug, write a feature, or refactor a file, and it works inside your repo.
  • Big jumps on coding tests. SWE-bench scores went up substantially over Claude 3.5.

If you want more

Worth knowing~30s
  • Claude Code's billing surprises. Multi-step tasks ran up real money fast — early users posted bills in the hundreds for a single project.
  • Extended thinking isn't always better. For everyday questions it's slower without much accuracy gain. Use it for hard things only.
Who should care~20s

Developers using Claude through Cursor, Zed, or the API. Teams evaluating AI coding assistants. Anyone tired of Cursor's pricing trying a CLI alternative.

What to do about it~20s

If you write code daily, install Claude Code and try it on a small task. Set a spend limit before you start — costs scale with thinking time.

Honest take~45s

Claude 3.7 was the moment Anthropic firmly took the coding crown from OpenAI. The extended-thinking toggle was a smart middle ground between always-fast and always-slow. The real story was Claude Code: a quiet declaration that the future of AI development isn't a fancy editor — it's a terminal where you tell the AI what to do and it does it. That bet has played out in their favour through the rest of 2025.

Other recent model updates

Sources

Last verified · 2026 · 05 · 05 · Found a fact wrong? corrections@aguidetocloud.com