Skip to main content

Roo Code 3.28.10 Release Notes (2025-09-29)

This release adds support for Claude 4.5 Sonnet across all providers, fixes GPT-5 compatibility with LiteLLM, and includes UI improvements.

Roo Code v3.28.10 Release

Claude 4.5 Sonnet Support​

We've added support for Anthropic's latest Claude 4.5 Sonnet model across all Claude-supporting providers (#8368).

According to Anthropic's announcement, Claude 4.5 Sonnet is:

  • State-of-the-art on SWE-bench Verified, maintaining focus for more than 30 hours on complex, multi-step tasks
  • Showing a significant leap forward on computer use with 61.4% on OSWorld benchmark (up from 42.2% just four months ago)
  • Delivering substantial gains in reasoning, math, and domain-specific knowledge across finance, law, medicine, and STEM
Claude 4.5 Sonnet model selection

The model is now available in the model selection dropdown for all supported providers.

Bug Fixes​

  • AWS Bedrock Claude Sonnet 4.5: Corrected model identifier for Claude Sonnet 4.5 on AWS Bedrock (thanks sunhyung!) (#8371)
  • GPT-5 LiteLLM Compatibility: Fixed GPT-5 models failing with LiteLLM provider by using correct max_completion_tokens parameter (thanks lx1054331851!) (#6980)

QOL Improvements​

  • Chat Icon Sizing: Chat interface icons now maintain consistent size in limited space (#8343)

Misc Improvements​

  • Telemetry Tracking: Enhanced analytics to track when users change telemetry settings (#8339)
  • Website Testimonials: Updated website with enhanced testimonials section (#8360)
  • Contributor Recognition: Improved contributor badge workflow with automated cache refreshing (#8083)