Gemini-3.7-flash is increadably slow

Hi,

You might not believe this.. but I kinda like gemini-3.7-flash.
I find it does quite decent code reviews etc.. especially at the price it is offered.

But, for some reason, in Cursor it is incredibly slow. Even a fairly simple question over a fairly simple codebase takes ages.

If I call the model programmatically via its api (unrelated to cursor), i get quite fast responses, I and have some long conversation history and rather long prompts.

So what I am saying is .. does anyone else have the same experience; Can something be done about it?

I really wanna use this models, but the sheer speed prevents me from using it more frequently.

Many thanks

Hey, thanks for the report. There’s one nuance here: a request in Cursor and a direct API call aren’t quite the same. In Cursor, extra context gets added to the prompt, like your codebase context, the system prompt, and tools. So the input token count and latency can be noticeably different from a clean API call, even if the question looks simple.

But to figure out if anything else is slowing it down, could you share:

  • The Request ID for the slow request right chat menu top-right > Copy Request ID. Note that if Privacy Mode is on, the Request ID may be unavailable.
  • Your Cursor version and OS.
  • Roughly how large the thread/context is when it starts getting slow.

One more quick test: run Cursor Settings > Network > Run Diagnostics. If you’re behind a corporate proxy or VPN, try enabling Disable HTTP/2 App Settings > search HTTP/2. Sometimes this noticeably affects streaming speed.

With the Request ID, I can check exactly where the time is going. Let me know how it goes.

Hi,

Thanks for the reply!

Yeah, i fully appreciate that using a model via Cursor is totally different compared to using it directly via its api. However, i have been (and still am) quite happy with how fast models are via cursor, and all the Opuses and Sonnets and GPTs, feel just as fast and often even faster compared to other in other harnesses i’ve tried or to calling them directly.

Even if I compare to older gemini models, gemini-3.7-flash is noticeably slower, which made open this discussion.

In any case, here are a couple of request ids:

  • 625fd467-3e11-431a-a59d-03b10a1aa9c9
  • d568be32-dec3-4b56-b6ee-a3235e475302

Yes, i have some legacy privacy mode, but I hope you will get some telemetry at least.

Also, I am talking about, first message, not a longer conversation. Immediately on the first message, i experience significant slowness compare to any other model (including past gemini’s).

I ran the network diagnostic: all green (i can share the logs if it needed)

This is cursor info:

Version: 3.17.8
VS Code Extension API: 1.128.0
Commit: 2fdd31c9f33f7fbe501f2d57772dc5bf64b63620
Date: 2026-08-20T02:18:12.724Z
Layout: IDE
Build Type: Stable
Release Track: Default
Electron: 40.10.3
Chromium: 144.0.7559.236
Node.js: 24.15.0
V8: 14.4.258.32-electron.0
xterm.js: 6.1.0-beta.291
OS: Darwin arm64 25.6.0

I hope this is enough info, i would be very grateful if you would look into this.

Cheers,
J.