Geeps Version 2.0

New features

• Rebuilt appearance settings and themes
  • Use a new theming engine inspired by the Codex and ChatGPT desktop apps.
  • Set any combination of background, foreground, and accent colors for light and dark modes. Theme saving will come in the next release.
  • Use different fonts for general chat content and code blocks. You can even use Papyrus — at your own risk.
  • Use a generic Compose button instead of provider logos.
  • Limit the chat width on Mac and iPad.
  • Send messages with the Enter key when the option is enabled in Settings.
• Image generation and editing with OpenRouter (requires Geeps Plus)
  • Load image models and their supported parameters directly from OpenRouter.
  • Choose OpenAI or OpenRouter as the default image provider.
  • Set image model and parameters per chat.
• Inline image generation with OpenAI (requires Geeps Plus)
  • Use the server-side image generation tool with models in the gpt-5 family and newer.
  • Receive generated images directly in text chats and continue editing them in the same conversation.
• Provider-neutral tool use and tool call display in chat (requires Geeps Plus)
  • Read web page securely extracts reader-friendly content from web pages. The tool uses the swift-readability package, which uses @mozilla/readability.
  • Exa Search adds web search to any provider. The tool uses the same Exa Search API that powers web search on OpenRouter and requires an Exa API key.
  • Set default search service (server side search vs Exa)
  • More tools, search services and Model Context Protocol (MCP) support are coming soon.
• iOS Share Sheet extension
  • Share a URL from Safari, then ask the model to read the web page, summarize it, or answer questions about it.
  • Share images or files to attach them to a new chat.
• Responses API support for custom providers
• Refreshed design across the app
• On-device conversation titles generated with Apple Foundation Models on supported devices.

Improvements and bug fixes

• Token usage is now displayed in the chat model menu when the API provides it.
• Added a Retry action to quickly retry a request after an API error.
• Added an option to show the provider and model under the title in the conversation list.
• Added provider logos and the option to change the icon for a built-in provider in addition to custom providers.
• Added a setting to show or hide reasoning output and tool calls.
• Added granular loading states for response phases such as Reasoning and Using a tool.
• Message actions now appear on hover on macOS or under an … button on iOS.
• Updated the message layout to use bubbles for user messages.
• Improved Markdown rendering, including better tables and typography.
• Improved reasoning rendering for OpenAI-compatible Chat Completions providers such as DeepSeek.
• Improved model loading for OpenAI-compatible providers. Geeps now handles more /models endpoint responses that do not fully match the OpenAI format.
• Significantly improved streaming and Markdown rendering performance for smoother scrolling and better performance in long conversations.

Many other smaller fixes and improvements under the hood.