Skip to main content

Roo Code 3.25.7 Release Notes (2025-08-05)

This release introduces Groq's GPT-OSS models, adds support for Claude Opus 4.1, brings two new AI providers (Z AI and Fireworks AI), and includes numerous quality of life improvements.

Groq GPT-OSS Models​

Groq now offers OpenAI's GPT-OSS models with impressive capabilities (#6732):

  • GPT-OSS-120b and GPT-OSS-20b: Mixture of Experts models with 131K context windows
  • High Performance: Optimized for fast inference on Groq's infrastructure

These models bring powerful open-source alternatives to Groq's already impressive lineup.

Z AI Provider​

Z AI (formerly Zhipu AI) is now available as a provider (thanks jues!) (#6657):

  • GLM-4.5 Series Models: Access to GLM-4.5 and GLM-4.5-Air models
  • Dual Regional Support: Choose between international and mainland China endpoints
  • Flexible Configuration: Easy API key setup with regional selection

📚 Documentation: See Z AI Provider Guide for setup instructions.

Claude Opus 4.1 Support​

We've added support for the new Claude Opus 4.1 model across multiple providers (#6728):

  • Available Providers: Anthropic, Claude Code, Bedrock, Vertex AI, and LiteLLM
  • Enhanced Capabilities: 8192 max tokens, reasoning budget support, and prompt caching
  • Pricing: $15/M input, $75/M output, $18.75/M cache writes, $1.5/M cache reads

Note: OpenRouter support for Claude Opus 4.1 is not yet available.

QOL Improvements​

  • Multi-Folder Workspace Support: Code indexing now works correctly across all folders in multi-folder workspaces (thanks NaccOll!) (#6204) - Learn more
  • Checkpoint Timing: Checkpoints now save before file changes are made, allowing easy undo of unwanted modifications (thanks NaccOll!) (#6359) - Learn more
  • Redesigned Task Header: Cleaner, more intuitive interface with improved visual hierarchy (thanks brunobergher!) (#6561)
  • Consistent Checkpoint Terminology: Removed "Initial Checkpoint" terminology for better consistency (#6643)
  • Responsive Mode Dropdowns: Mode selection dropdowns now resize properly with the window (thanks AyazKaan!) (#6422)
  • Performance Boost: Significantly improved performance when processing long AI responses (thanks qdaxb!) (#5341)
  • Cleaner Command Approval UI: Simplified interface shows only unique command patterns (#6623)
  • Smart Todo List Reminder: Todo list reminder now respects configuration settings (thanks NaccOll!) (#6411) - Learn more
  • Cleaner Task History: Improved task history display showing more content (3 lines), up to 5 tasks in preview, and simplified footer (#6687)
  • Internal Architecture: Improved event handling for better extensibility (#6606)

Provider Updates​

  • Fireworks AI Provider: New provider offering hosted versions of popular open-source models like Kimi and Qwen (thanks ershang-fireworks!) (#6652)
  • Cerebras GPT-OSS-120b: Added OpenAI's GPT-OSS-120b model to Cerebras provider - free to use with 64K context and ~2800 tokens/sec (#6734)

Bug Fixes​

  • Mode Name Validation: Prevents empty mode names from causing YAML parsing errors (thanks kfxmvp!) (#5767)
  • Text Highlight Alignment: Fixed misalignment in chat input area highlights (thanks NaccOll!) (#6648)
  • MCP Server Setting: Properly respects the "Enable MCP Server Creation" setting (thanks characharm!) (#6613)