v4.0.12 v4.0.13

medium risk

farion1231/cc-switchOct 14, 2026 → Oct 15, 20261 day between releases

What's New

Introduces unified connection modes for Claude Desktop and caps GPT context window at 272K.

Breaking Changes

  • Claude Desktop users with model mapping cards automatically switch to Route mode.
  • Codex 'Use classic tools for sub-agents' setting is now enabled by default.

New Features

  • Unified Direct, Routed, and Aggregated modes for Claude Desktop.
  • Codex allows selecting a specific model for auto-approval.
  • Codex aggregation mode allows sub-agents separate models and inference intensities.
  • Claude Code Haiku profile can declare 1M context.
  • New presets: TokenFactory and Modelflare.

Improvements

  • GPT model context windows capped at 272K to prevent upstream rejections.
  • Windows portable update button now downloads zip directly.
  • Fixed configuration saving errors (duplicate key).
  • Fixed Windows 11 window freezing and high CPU usage.

Bug Fixes

  • Fixed Codex saving configuration 'duplicate key' errors.
  • Fixed Codex remote compression failure in routed mode.
  • Fixed Windows 11 window freezing and high CPU usage.

Upgrade Notes

Restart Claude Desktop after switching connection modes. Existing model mapping cards will automatically switch to Route mode. Codex 'Use classic tools for sub-agents' is now enabled by default.

The Two Releases

v4.0.12

CC Switch v4.0.12

Oct 14, 2026· 8 days ago

CC Switch v4.0.12 introduces full management capabilities for tools on the Applications page, including uninstalling specific installations and verifying Node versions before upgrades. It adds a new 'Local' group for local inference services and 14 new cloud provider presets. Additionally, the release optimizes resource usage by extending the session log scan interval and adds Turkish language support.

Full release notes
v4.0.13

CC Switch v4.0.13

Oct 15, 2026· 7 days ago

This release introduces unified Direct, Routed, and Aggregated connection modes for Claude Desktop, improves GPT context window management by capping it at 272K, and adds an auto-approve model feature for Codex. It also includes fixes for configuration saving errors and Windows 11 window freezing.

Full release notes