Follow Discord

DisclaimerUnofficial, and not affiliated with Anthropic. Nearly all of this is read straight out of what ships: npm bundles, captured prompts, published docs. Anthropic's own notes go in verbatim, marked as theirs. The rest is my reading, and every entry carries the strings behind it. If one looks wrong, vote it down and say why.

← All of v2.0.5 Home All releases ←olderv2.0.3 v2.0.8newer→

Claude Code v2.0.5 2 Oct 2025

Extended Thinking Support for Sonnet 4.5

What: Claude Code now supports extended thinking mode when using Claude Sonnet 4.5, enabling more transparent reasoning for complex tasks.

How to use:

# Enable thinking via environment variable
MAX_THINKING_TOKENS=10000 claude

# Or configure in your settings file
# Set alwaysThinkingEnabled: true in your Claude Code settings

Details:

Only available when using Claude Sonnet 4.5 model
Can be controlled via MAX_THINKING_TOKENS environment variable (set to any positive integer to enable)
Can be configured permanently via alwaysThinkingEnabled in settings
Subject to Statsig feature gate thinking_on_default for gradual rollout
Provides visible reasoning steps during Claude's problem-solving process
Evidence: hL1() at line 364041 in v2.0.5

See this entry in the whole of v2.0.5 →