Skip to main content

Roo Code 3.33 Release Notes (2025-11-18)

Roo Code 3.33 introduces support for Gemini 3 Pro Preview and native tool calling for OpenAI-compatible providers, along with improved Gemini defaults and numerous native tool protocol fixes.

Major Features​

Gemini 3 Pro Preview​

Roo Code now supports Google’s Gemini 3 Pro Preview model through direct Gemini, Vertex AI, and aggregator providers like OpenRouter and Requesty (#9357, #9327):

  • 1M-token, reasoning-capable model: Handles very large conversations while providing higher-quality multi-step reasoning on complex coding and refactoring tasks.
  • Strong eval performance: Achieves a 100% score on internal Roo Code evals and 76.2% on SWE-bench Verified, giving more consistent solutions on real-world coding tasks.
  • Reliable tool usage: Executes complex multi-step tool workflows without getting stuck or losing track, especially in long, tool-heavy tasks.
  • Better out-of-the-box defaults: Uses gemini-2.5-pro by default where supported, sets a more natural temperature of 1, cleans up the Gemini model list, and includes reasoning / “thought” tokens in cost reporting so usage numbers better match provider billing.
  • Image Preview Support: Added support for image previews with Google Gemini 3 Pro image generation models (#9440).

QOL Improvements​

  • Native Tool Calling for OpenAI-compatible Providers (#9369): Enables native tool use with a wider range of models and providers that support the OpenAI format, offering more robust tool interactions in both streaming and non-streaming modes.

  • Minimax Provider Support (#9455): Added support for Minimax as an Anthropic-compatible provider.

  • Reasoning History (#9451): Reasoning is now stored in conversation history for all providers.

  • Improved Read File Tool (#9422): Enhanced the read_file tool description with clearer examples.

  • Git status in environment details (#9310): Shows git status information in environment details so agents have more context about untracked, modified, and staged files.

  • Tool protocol selector in advanced settings (#9324): Lets you choose which tool protocol to use (such as XML vs native) without editing config files.

  • Dynamic tool protocol resolution (#9286): Resolves the active tool protocol using a clear precedence hierarchy.

  • Improved Modes view toolbar (#9077): Moves Import/Export into the Modes view toolbar and cleans up the Mode edit view.

  • Cloud agent CTA points to setup page (#9338): Updates the cloud agent call-to-action to link directly to the setup page.

  • Roo Code Cloud provider pricing page (#9195): Adds a pricing page and related Cloud provider tweaks.

  • Gemini native tool calling (experimental, Google Gemini provider only) (#9343): Adds an experimental native tool option (off by default) for the Google Gemini provider so Gemini-based workflows can use tools without XML translation and behave more like other native-tool providers.

  • Roo Code Cloud credit balance in the extension (#9386): Shows your Roo Code Cloud RCC credit balance directly in the Roo provider settings so you can track usage and avoid unexpected interruptions.

  • Explicit 0 pricing for free models (#9419): Displays explicit 0 input/output token prices for free models in the pricing UI so it is always clear which options are free.

Bug Fixes​

  • Native Tool Protocol Race Condition (#9363): Fixed a race condition where switching profiles or changing API configuration could cause 400 errors.
  • Native Tool Repetition Detection (#9377): Fixed false positive "stuck in a loop" errors when using native tools repeatedly.
  • Strict Native Protocol Parsing (#9371): Ensured no XML parsing occurs when the native protocol is active.
  • Gemini Reasoning Config (#9375): Fixed an issue where high/low reasoning effort configuration was not being applied correctly.
  • Gemini Thought Validation (#9380): Resolved integration errors with Gemini reasoning models.
  • Prevent duplicate tool_result blocks in native protocol (#9248): Ensures each native tool call emits a single tool_result block.
  • Format tool responses for native protocol (#9270): Normalizes the structure of tool responses for easier model comprehension.
  • Centralize toolProtocol configuration checks (#9279): Uses a single source of truth for toolProtocol configuration.
  • Preserve tool blocks in conversation history (#9319): Keeps native protocol tool blocks intact in history.
  • Prevent infinite loops after successful finalization (#9325): Fixes a regression where certain native tool flows could loop.
  • Sync parser state with profile and model changes (#9355): Keeps the conversation parser aligned with the active profile and model.
  • Pass tool protocol to truncation errors (#9358): Ensures truncation errors know which tool protocol is active.
  • Structured JSON Returns (#9373): Updated tools to consistently return structured JSON when using the native protocol.
  • VS Code theme-colored outline button borders (#9336): Aligns outline button borders with the current VS Code theme.
  • Use shields.io badges instead of badgen.net (#9318): Replaces broken badge URLs with shields.io.
  • Cap git status file sampling in evals (#9322): Adds a maximum for git status files in eval settings.
  • Preserve images in native tool call results (#9401): Ensures images attached to messages are preserved and passed through native tool calls so follow-up reasoning can still use visual context instead of losing it as plain text only.
  • Faster router model metadata loading (#9410): Reduces redundant getModel() calls and adds a disk-backed metadata cache so router-based providers start faster and are more reliable when loading model info.
  • Environment Details Duplication (#9442): Fixed an issue where environment_details were duplicated when resuming cancelled tasks.
  • Preserve Reasoning Flag (#9453): Fixed the preserveReasoning flag to correctly control API reasoning inclusion.
  • OpenAI Parallel Tool Calls (#9433): Prevented OpenAI Native parallel tool calls when using native tool calling to avoid errors.
  • Search and Replace Parsing (#9456): Improved symbol parsing in the search and replace functionality.
  • Skipped Tool Results (#9457): Fixed an issue where tool_result blocks were not sent for skipped tools in the native protocol.
  • Markdown Formatting (#9458): Improved markdown formatting and added reasoning support.