Skip to main content

Roo Code 3.36.3 Release Notes (2025-12-10)

This release adds browser screenshot saving, xhigh reasoning effort support, native tools default for OpenRouter, plus provider updates and bug fixes.

Roo Code v3.36.3 Release

Browser Screenshot Saving​

The browser tool now supports saving screenshots to a specified file path with a new screenshot action, enabling you to capture visual state during browser automation tasks. (#9963)

Extra-High Reasoning Effort​

Users of the gpt-5.1-codex-max model with the OpenAI provider can now select "Extra High" as a reasoning effort level, enabling maximum reasoning capability for complex tasks (thanks andrewginns!) (#9900)

OpenRouter Native Tools Default​

OpenRouter models that support native tools now automatically use native tool calling by default, improving tool calling reliability without manual configuration. (#9878)

QOL Improvements​

  • Unified Context-Management UX: Real-time feedback for context operations with truncation notifications and condensation summaries (#9795)
  • Better OpenAI Error Messages: Enhanced error handler extracts detailed information from API errors for easier troubleshooting (#9639)
  • Token Counting Optimization: Removed separate API calls for token counting, improving performance (#9884)
  • Tools Decoupled from System Prompt: Tool-specific instructions are now self-contained in tool descriptions (#9784)

Bug Fixes​

  • Tool Protocol Selector: Always show tool protocol selector for OpenAI-compatible providers (thanks bozoweed!) (#9966)
  • apply_diff Filtering: Properly exclude apply_diff from native tools when diff is disabled (thanks denis-kudelin!) (#9920)
  • API Timeout Handling: Fixed disabled API timeout (set to 0) causing immediate request failures (thanks dcbartlett!) (#9960)
  • Reasoning Effort Dropdown: Respect explicit supportsReasoningEffort array values and fix disable handling (#9970, #9930)
  • Actual Error Messages: Display actual API error messages instead of generic text on retry (#9954)
  • Stream Hanging Fix: Process finish_reason to emit tool_call_end events for OpenRouter, Roo, and xAI providers (#9927, #9929)
  • tool_result ID Validation: Validate and fix tool_result IDs before API requests to prevent Anthropic rejections (#9952)
  • Suppressed Internal Error: Fixed 'ask promise was ignored' error leaking to LLM conversations (#9914)
  • Provider Sanitization: Fixed infinite loop when using removed/invalid API providers (#9869)
  • Context Icons Theme: Context-management icons now use foreground color to match VSCode themes (#9912)
  • Eval Runs Deletion: Fixed foreign key constraint preventing eval run deletions (#9909)
  • OpenAI-Compatible Timeout Reliability: Added timeout handling to OpenAI-compatible providers to prevent indefinite hangs (thanks dcbartlett!) (#9898)

Provider Updates​

  • DeepSeek V3.2: Updated to V3.2 with 50% price reduction, native tools enabled by default, and 8K max output (#9962)
  • xAI Models Catalog: Updated catalog with corrected context windows, image support for grok-3/grok-3-mini, and removed deprecated models (#9872)
  • xAI Tool Preferences: Configured xAI models to use search_replace for better file editing compatibility (#9923)
  • DeepSeek V3.2 for Baseten: Added DeepSeek V3.2 model support (thanks AlexKer!) (#9861)
  • Baseten Model Tweaks: Improved maxTokens limits and native tools support for stability (#9866)
  • Bedrock Models: Added Kimi, MiniMax, and Qwen model configurations (thanks jbearak!) (#9905)

Misc Improvements​

  • Announcement UI Updates: Improved announcement visuals with updated social icons and GitHub stars CTA (#9945)
  • Improved Error Logging: Better error context for parseToolCall exceptions and cloud job errors (#9857, #9924)
  • search_replace Native Tool: New tool for single-replacement file operations with precise targeting via unique text matching (#9918)
  • Versioned Settings Support: Internal infrastructure for API-side versioning of model settings with minimum plugin version gating (#9934)
  • OpenRouter Telemetry: Added API error telemetry for better diagnostics (#9953)
  • Evals Streaming Stats: Tool usage stats now stream in real-time with token usage throttling (#9926)