Geeps Version 1.6

March 2026

Release 1.6

Support for OpenAI Responses API, recommended for use with reasoning models like gpt-5 and later and supporting gpt-codex models.
- Support for reasoning_effort and verbosity parameters for OpenAI (configure defaults in API Settings > Model parameters, or set for individual chats)
Support for Claude adaptive thinking settings (available for Sonnet and Opus 4.6+)
Support for new automatic prompt caching for Claude models (also supported on OpenAI Responses API)
Thinking / Resoning Summary displayed in chat when reasoning effort / adaptive thinking is enabled
Improved web search across providers
- OpenAI Responses API uses the native web_search tool and Chat Completions uses the newer gpt-5-search-api model for search
- Search results are now displayed in a sheet / popover when tapping / clicking the Sources button
  - OpenRouter web search now uses web plugin instead of appending :online to model (behavior is the same)

Chat settings moved to a sheet accessible via the + button (quick select for prompt and model parameters)
Improved model parameters UI in API Settings
Tap on file attachment in chat to see the extracted text content in a sheet
Markdown rendering improvements:
- thematic break (---) support
- improved line break handling
- better table column rendering with vertical dividers
- inline code has a subtle background
OpenRouter free models filter is now preserved
Fixed an issue where provider settings would not sync across devices correctly because of duplicates in iCloud
Fixed accent colors for some UI controls
OpenAI models with with timestamps are filtered out from the list
Other minor fixes and improvements