Anthropic Releases Claude 3.7 Sonnet With Reasoning and Claude Code CLI
Anthropic launches Claude 3.7 Sonnet, its first reasoning model, alongside Claude Code, an agentic command-line tool for software development. The new model matches the pricing of its predecessor while offering extended output capabilities.
Anthropic releases Claude 3.7 Sonnet today, skipping the 3.6 designation because users already adopted that name for an earlier update. This new model matches the $3 per million input tokens and $15 per million output tokens pricing of Claude 3.5 Sonnet while introducing Anthropic's first official reasoning capabilities. The model joins similar thinking models from OpenAI, Google, DeepSeek, and others in a growing industry trend.
The standout feature of Claude 3.7 Sonnet is its extended thinking mode, which allows the model to work through complex problems before delivering a final answer. It supports up to 64,000 output tokens by default, but developers can enable a special beta header to unlock up to 128,000 output tokens. This expanded capacity makes the model particularly useful for intricate code generation and comprehensive content creation.
Alongside the new model, Anthropic previews Claude Code, a command-line interface tool that acts as an autonomous coding agent. Claude Code operates directly in the terminal, reads and modifies files, and executes commands to run tests and iterate on code. This combination of a powerful reasoning model and an agentic CLI provides developers with a robust new environment for software engineering tasks.