Skip to main content

Roo Code 3.31.3 Release Notes (2025-11-13)

This patch fixes OpenAI native encrypted reasoning handling, corrects Anthropic input token normalization, and cleans up context management terminology.

Roo Code v3.31.3 Release

Bug Fixes​

  • OpenAI native encrypted reasoning parameters: Restores usability of gpt-5-chat-latest by including encrypted reasoning only when a reasoning effort is set and enforcing medium-only verbosity to prevent invalid parameter errors (thanks politsin!) (#9231)
  • Roo Anthropic token normalization: Prevents double-counting cached tokens so usage metrics, cost estimates, and context calculations are accurate across providers (#9224)

Misc Improvements​

  • Terminology cleanup for context handling: Rename "sliding window" to "context management" and unify the code path under a single entry point, manageContext, for clearer behavior and documentation alignment (#9206)