Gemini Issues and Errors

I have endured two months of frustration using the gemini-2.5-pro-exp-03-25 model on your platform, and this charade ends now. As of today:

  • Model Non-Responsiveness: I no longer receive any code outputs—only abstract “plans” with zero execution.
  • Connection Failures: Countless requests fail with

“We’re having trouble connecting to the model provider. This might be temporary – please try again in a moment. (Request ID: 81e321ea-8890-4da8-8cbc-8817f7a58689)”

Paying for a half-baked service that consistently breaks is unacceptable.

Immediate Actions Required:

  1. Restore gemini-2.5-pro functionality within 24 hours.
  2. Provide a concrete timeline for eliminating the “trouble connecting” errors—permanently.
  3. Issue a full refund for any time lost while your service was down.

Failure to address these demands will force me to:

  • Cancel my subscription without hesitation.
  • Share my experience across developer communities and social media.
  • Switch to Copilot or any other reliable alternative that actually delivers on its promises.

This is your final opportunity to prove Cursor can live up to its claims.

1 Like

Hi, thanks for the post.

Google has been seeing some degradation in the past hour or so, which means their ability serve Gemini has also been degraded.

We believe this has started to recover now, but as Gemini is a closed-source model, we cannot host it ourselves and must rely on providers like Google to serve the model to us.

You always have the other models within Cursor available to you at all times, but feel free to email us at [email protected] if you would like a refund or to cancel your plan.

Hopefully things are better moving forward.

I have tried others models as well and they are failing as well. Please check the attached screenshot.

Also I haved seen any incident on google’s end, could you provide the source for your information?

We are in direct contact with Google while they get this resolved, so this issue has come from them directly.

Unfortunately, I cannot share the details of that conversation, so I can’t help much if you don’t believe me for whatever reason.

Regarding your screenshot, it looks like the LLM may have hallucinated or got confused on how to call the tools it can use in Cursor - starting a new chat should resolve that.

You really should look at refunding credits for failed calls - especially these that didnt register as failed calls when they stopped half way through.

This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.