Cursor IDE Opus 4.6 model response speed severely degrades as API usage approaches 90% on Ultra plan

Support Bug Reports
,
1

Where does the bug appear (feature/product)?

Cursor IDE

Describe the Bug

I am a Cursor Ultra plan subscriber who exclusively uses the Opus 4.6 model for all tasks. I have observed a severe, progressive degradation in response speed as my API usage percentage increases, while other models remain fully responsive.

Steps to Reproduce

  1. Subscribe to Cursor Ultra plan and use Opus 4.6 as the primary model for all development tasks.
  2. Monitor API usage percentage in the Cursor dashboard.
  3. Observe response speed as usage approaches 90%, then 95%, and near 100%.
  4. Compare response speed by switching to other available models at the same usage level.

Operating System

MacOS

Version Information

Version: 2.6.22
VSCode Version: 1.105.1
Commit: c6285feaba0ad62603f7c22e72f0a170dc8415a0
Date: 2026-03-27T15:59:31.561Z
Build Type: Stable
Release Track: Default
Electron: 39.8.1
Chromium: 142.0.7444.265
Node.js: 22.22.1
V8: 14.2.231.22-electron.0
OS: Darwin arm64 25.4.0

Does this stop you from using Cursor

No - Cursor works, but with this issue

5

Hey, a few users are reporting something similar. There’s a related thread: Has anyone noticed performance issues with Opus 4.6 (Max Mode) recently?

A couple things to keep in mind:

  • Opus 4.6 with extended thinking is naturally slower than other models because it does extra reasoning steps.
  • There were some recent issues on the Anthropic API side: Claude Status - Elevated errors on requests to Claude Opus 4.6, which could have affected speed.
  • The team is aware of the Opus speed reports. No specific ETA yet, but every report helps us prioritize.

To look into your specific requests, we’ll need Request IDs from slow sessions. Top right of the chat > Copy Request ID. 2 to 3 examples is enough.

Also, are you using Opus in Max Mode or normal mode? About how long are replies taking, like 30 seconds or 1 minute+?

As a workaround, try starting fresh chats instead of continuing long threads. Large accumulated context can add a lot of latency.