Skip to content

[Bug] Gemini 3 Pro Preview fails with "Error processing response" (Paid Tier/Reasoning Issue) #247

Closed

Description

@dm2mymcszt-commits

Hi dev, first of all, I love you. Thank you for this app, you ate that up.

I am using the gemini-3-pro-preview model with a Paid API Key (Pay-as-you-go Tier).
When attempting to run any tool on the Windows version of the app, it instantly fails with the error message:

"An error occurred while processing the response."

Context & Debugging

  • API Status: My API key is valid and paid. I have confirmed this by testing the exact same key and model on the iOS port (ProseKey AI), where it works perfectly.
  • Model: gemini-3-pro-preview (Custom Model entry).
  • Comparison: If I switch the Windows app to gemini-3-flash-preview, it works fine.
  • Suspected Cause: Gemini 3 Pro enforces "Reasoning/Thinking" blocks in its JSON response. The Windows client appears to be choking on these thought signatures (parsing error), whereas the iOS version handles/strips them correctly.

To Reproduce

  1. Enter a valid Paid API Key in Settings.
  2. Set the Custom Model to gemini-3-pro-preview.
  3. Select text and trigger any tool.
  4. Observe the "An error occurred while processing the response." message.

Environment

  • OS: Windows 10
  • App Version: latest

Activity

  1. theJayTea commented on Jan 13, 2026

    @theJayTea
    Owner

    Hi @dm2mymcszt-commits! I see this is certainly something I can fix.

    I'm curious -- what's your use case for paid Gemini 3.0 Pro with Writing Tools?

    Smaller LLMs have been remarkably great at most of what I've seen people using Writing Tools for (summarising, proofreading, text manipulation...), and I'd imagine that one would prefer the speed of the smaller models too [in addition to the free/cheap price :P].

    Curiosity aside, I can still look into adding 3.0 Pro compatibility in the future :)

  2. dm2mymcszt-commits commented on Jan 13, 2026

    @dm2mymcszt-commits
    Author

    Since I'm already paying for the API tier I tend to default to the strongest model available just to get the most out of my subscription lol..

    While I do use the text tools I also use the Ask AI feature heavily for complex or basic questions because I am 2 lazy to open Firefox..

    That said u are absolutely right about the speed because I gave Gemini 3 Flash a try and the latency difference is massive it definitely feels snappier for general proofreading/editing task
    Please feel free to treat this as low priority for now I'll stick to Flash for the speed benefits!!

    If Pro compatibility gets added in the future that would be like a little bonus but it's definitely not blocking me from enjoying the app. Thanks again love u !

  3. theJayTea commented on Jan 13, 2026

    @theJayTea
    Owner

    Cool! I'll look into adding this in the future, as right now I have a good couple exciting things I'm working on for Writing Tools (improved stability, per-button hotkeys, more polish, etc!).

    I'll close this issue for now, but I'll keep this in mind :)

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions