XIO

xWire

Posts from every feed QUEST has found across the web — blogs, news, podcasts, channels and more.

Cursor - Community Forum - Topics tagged grok

Topics tagged grok

ConnectError: [internal] internal error

Where does the bug appear (feature/product)? Cursor IDE Describe the Bug Bug Description I am encountering a persistent ConnectError: [internal] internal error in Cursor. The issue began immediately after resuming my laptop from an extended sleep state (approx. 17 hours). Submitting any prompt immediately fails with an internal server error, making the agent and chat unusable with this model. Error Trace & Request ID Plaintext Request ID: 43b081b6-08ab-47b8-b852-2babdc1fcc9f [internal] internal…

[Bug] Agent encountered an unexpected error during execution (Model: Grok 4.7 High Fast)

Where does the bug appear (feature/product)? Cursor IDE Describe the Bug While using the Agent feature in the Cursor IDE (using the Grok 4.7 High Fast model), the run suddenly failed. It was processing a request and showing file changes (9 files changed), and then a red error box appeared at the bottom saying: “Agent encountered an error. This run hit an unexpected error. Retry or send a follow-up below.” Clicking “Retry” does not seem to fix the issue permanently. Steps to Reproduce Open…

Grok 4.7 is not found

Where does the bug appear (feature/product)? Cursor IDE Describe the Bug Grok 4.7 is not found in the editor chat Steps to Reproduce Just open the chat and start talking to Grok 4.7 Operating System Linux Version Information Version: 3.27.22 VS Code Extension API: 1.128.0 Commit: b1aa50cb6db84fd3e8c808259cb172ea73425c10 Date: 2026-10-27T00:33:02.045Z Layout: IDE Build Type: Stable Release Track: Default Electron: 42.11.16 Chromium: 148.0.7778.280 Node.js: 24.19.0 V8: 14.8.178.38-electron.0…

Why is Grok 4.7 Extra High so incredibly dumb?

Where does the bug appear (feature/product)? Cursor IDE Describe the Bug Grok can’t even handle the most basic tasks. Steps to Reproduce I had five “color selection” elements; Grok broke the design for three of them on its own—even though I never asked it to change the design at all. I asked it to fix the issue, and it ended up breaking the design for the first two as well. I asked it to revert the buttons to how they looked before—instead, it came up with an entirely new design of its own. And…

Grok is slow :(

Grok 4.7 has become noticeably slower over the past few weeks Has anyone else noticed that Cursor has become significantly slower when using Grok 4.7 over the past few weeks? What makes this particularly frustrating is that Grok 4.7 seems noticeably slower than other models, even though those models consume API quota while Grok is included. What is going on? Is this a known issue, a capacity problem, or something else? I’m the only person among my peers who hasn’t switched to Claude yet. I’ve…

Task under Auto rejects gemini-4-argon; under BYOK Gemini parent, grok-4.7-high-fast Task seats silently run as Gemini 4 Argon

Where does the bug appear (feature/product)? Cursor IDE Describe the Bug Chat model picker and Task model allow-list are different catalogs. Two related failures: Under Auto (or Grok) parent: Task only accepts a tiny Grok allow-list (observed: cursor-grok-4.6-high-fast, grok-4.7-high-fast). gemini-4-argon / gemini-4-argon-high are rejected with “Invalid model selection” even though Gemini 4 Argon is selectable as the parent chat model. Chat picker ≠ Task allow-list. Under Gemini 4 Argon parent…

Usage dashboard lists the same Grok models under both “Cursor Models” and “Other Models” without spillover labeling

Where does the bug appear? Cursor Web Dashboard → Usage → Included Usage breakdown Describe the Bug The Included Usage tables look mixed between Cursor Models and Other Models: The same model IDs appear in BOTH tables (grok-4.7-high-fast, cursor-grok-4.6-high-fast) with no label that the Other Models rows are spillover after the Cursor Models pool is exhausted. Naming is inconsistent: bare grok-4.7-high-fast / grok-4.7-high under Cursor Models, while cursor-grok-4.6-* appears under both (and…

Models ridiculously guardrailed post update

Hello everyone, I wanted to share my professional opinion regarding the latest Cursor IDE update and the changes in model behavior. I’m a security researcher and have been a Pro++ subscriber for almost a year. I regularly use Auto models, particularly Grok, for quick source code reviews focused on identifying potential security vulnerabilities. Most of this work involves my personal projects and publicly available source code, where I test different rules and skills. I generally don’t use…

Is Continuous RL good anymore?

Just wanna discuss on the Continuous RL on Grok and Composer 2.5. Sure we don’t know how Cursor team managed to filter out the quality of the data they feed to the model to do continuous RL, but things seem to go down the hill. Composer used to be good actually, but now it and Grok just keep ignoring instructions, having bad output quality, restate what the user says and called it done explaination… while still burn a lot of token while doing so. Token price on those models are very cheap I…

Will Composer and Grok ever improve?

A few months ago, I started using Cursor and was quite excited about the tool. However, I’ve found that the token costs are extremely high, causing my plan’s allowance to run out very quickly. Furthermore, models like Grok and Composer are inefficient. I also frequently run into code-editing issues where characters with diacritics—common in languages ​​like Portuguese—end up as “mojibake” (garbled text). In contrast, tools like Claude Code and Codex are far superior; they are much more…

Model choice in subagent sometimes not working appropriately

Where does the bug appear (feature/product)? Cursor IDE Describe the Bug In cursor settings – Models page, users can select task model for subagents via explore subagent model, but sometimes subagents are not following the current choice. Is it a bug? Steps to Reproduce By using other models like gpt5.6 in plan, when I specify grok4.7 as subagents model, sometimes it will use gpt5.6 instead. Operating System Windows 10/11 Version Information Version: 3.26.16 (user setup) VS Code Extension API:…

Can't use Other Models after maxing out Cursor Models

Where does the bug appear (feature/product)? Somewhere else… Describe the Bug In this billing period, I’ve only used grok 4.7 or grok 4.6 and my ‘Other Models’ usage was at 0%. On reaching 100% on ‘Cursor Models’, ‘Other Models’ has also shot to 100% without any usage and now I can’t use it. Is this expected? Steps to Reproduce Use only Cursor Models to 100% Expected Behavior Cursor Models and Other Models usage remains separate Operating System Windows 10/11 MacOS Version Information Not tied…

New agents reset to Grok 4.7 despite Composer 2.5 Fast being selected

Where does the bug appear (feature/product)? Cursor IDE Describe the Bug I’ve configured the Explore Subagent Model to Composer 2.5 Fast in Cursor Settings → Models. However, every time I create a new Agent, the model resets to Grok 4.7 instead of using my selected model. Expected behavior: New agents should respect the configured model preference. Actual behavior: New agents keep defaulting to Grok 4.7. Frequency: Every new Agent. Please investigate why the selected model preference isn’t…

Grok 4.7 Extra High quality is getting worse by the day

Where does the bug appear (feature/product)? Cursor IDE Describe the Bug Cursor’s native agent with Grok 4.7 Extra High has gotten significantly worse at regular coding tasks over the past few weeks. Things it could easily accomplish last month seem impossible now. Today, it couldn’t even fix a scrolling issue on a webpage after four attempts, something it would have handled without any trouble a month ago. Claude fixed it on the first try in under a minute with a one-line change. I’ve been…

Composer 2.5 regression + premium models burning credits for nothing

Composer 2.5 regression + premium models burning credits for nothing. Composer 2.5 has become noticeably worse. The quality regression is frustrating, especially for Android Studio and PHP development. The bigger issue is that premium models consume your usage like crazy without actually solving anything. Gemini, even the latest models, has been garbage in my experience. Grok 4.6/4.7 are equally disappointing for my workflow. A single task using Gemini 3.8 Flash consumed 70% of my $60 plan…

Custom DeepSeek API integration causes "Provider Error" on native Cursor models

Hi Cursor Team, I am experiencing an issue where configuring a custom API key for DeepSeek seems to break the native Cursor models. Environment: OS: Windows Mode: Ask and Agent Mode Cursor Version: Version 3.19.7 Steps to Reproduce: I added my DeepSeek API key in Settings > Models > API Keys. I enabled “Override OpenAI Base URL” and set it to https://api.deepseek.com/v1. I verified my DeepSeek key is valid by successfully calling https://api.deepseek.com/models via curl. I manually added the…

Max Mode Required for Grok 4.7 Fast

Where does the bug appear (feature/product)? Cursor IDE Describe the Bug When attempting to use Grok 4.7 Fast, I am required to be in Max Mode. I’m am on the X Subscription plan. When I read the docs, its saying that Max mode is for legacy only. Not sure if I am doing something wrong… Steps to Reproduce Trying to use Grok 4.7 Fast for any prompt. then says Max Mode required. Expected Behavior Max mode not required. Operating System MacOS Version Information Version: 3.25.9 (Universal) VS Code…

Very frustrated with how Grok models is ignoring rules now

Where does the bug appear (feature/product)? Cursor IDE Describe the Bug Grok models are very poor on instruction following now. Please take a look at the request ID provided below. I’ve sent a very straightforward request, and with all the applied rules, the Grok Agent is still prefer its own way of working and output, ignoring most if not all of my pre-defined rules. It sure can realize that it didn’t follow the rule when I asked for an explanation and fix it correctly after that, but only…

Please ask the official: Under the package limit, is there any difference in the total amount calculation when calling functions within Cursor compared to calling the API for Grok 4.7-high?

I am a Premium plan subscriber. I have a specific question regarding the usage quota consumption for the “Grok 4.7-high” model: Within the scope of my existing plan’s quota, is there any difference in the total quota consumed when accessing this model via the Cursor interface versus accessing it through the Grok API? If both methods draw from the same plan quota, are the token deduction rates identical? This is an important point for me, as it will determine whether or not I choose to use…

[Bug] All models fail with resource_exhausted on new conversations - Pro+ plan

Client: Cursor 3.24.18 (stable, updated Oct 14) Account: ahqonejewelry@gmail.com — Pro+ $60/mo Usage: Cursor Models 28% used, Other Models 0% used (quota NOT exhausted) Issue: All requests fail with resource_exhausted (capacity error), including a brand-new empty conversation with no attachments and a single short message. Request IDs: fd5bfa87-2043-474c-a5ef-7edc63d2a426 (modelId: grok-4.7) 437498d0-6394-49d9-9e70-fa2275966817 (modelId: claude-opus-4-8) Errors observed: grok-4.7: ConnectError…

Notion Cloud Agent requires prepaid credits even when Grok 4.7 is set and weekly pool has plenty left

Hi, I’m on The X Subscription Premium plan. My weekly included usage is only about 6% used (resets Oct 24). I set the default model for Cloud Agents to Grok 4.7 (256K, High, Fast). When I try to trigger a Cursor agent from Notion (via @Cursor in a comment or by moving a task to “Ready for agent”), I consistently get this error: “The connected Cursor account is out of included usage or needs a paid plan. Check the plan and spend limit in the Cursor dashboard at cursor.com/dashboard, then try…

Can't change Cloud Agents off Fast Mode – Eating my Premium limits!

Where does the bug appear (feature/product)? Cloud Agent (GitHub, Slack, Web, Linear) Describe the Bug I really wasn’t expecting this from Cursor. It looks like Cloud Agents are now defaulting to Grok Fast mode, and the web interface completely blocks me from switching it off. No matter what I click, it forces the Fast model execution. As a Premium plan user, this is incredibly frustrating. This bug is going to burn through my token limits and credit pool at lightning speed without my consent.…

Cursor CLI ACP reports Grok 4.6 Fast despite explicitly selecting the listed non-Fast model

Where does the bug appear (feature/product)? Cursor CLI Describe the Bug Cursor CLI lists cursor-grok-4.6-high as an available non-Fast model. However, when I explicitly select that model at startup, the ACP session/new response reports: grok-4.6[effort=high,fast=true] The model configuration’s currentValue also reports fast=true. I initially encountered this through OpenHands Agent Canvas, but I reproduced the same model metadata mismatch by launching Cursor ACP directly from PowerShell,…

What on earth are you guys doing with Grok 4.7!

Where does the bug appear (feature/product)? Cursor IDE Describe the Bug Is the rumor about Grok 4.7 severely draining tokens actually real? I’ve burned through two Cursor Ultra plans only to end up with an utterly disappointing result, with nowhere to even vent my frustration. Every single task takes over 60 minutes just to locate where the issue is, deep-“thinking” across every file in the project. It drains my Ultra tokens at a dizzying speed, yet ultimately fails to fix even a basic queue…

Grok 4.7 Weird Output

Where does the bug appear (feature/product)? Cursor IDE Describe the Bug Not sure if this is a bug or not but Grok 4.7 is just spewing this in the chat as it is thinking about resolving the issue that I prompted it to fix. This is just a snippet but it is just continuous. "The. Last. One. And. The. Last. One. Ends. Now. And. The. Call. Begins. After. It. And. The. Call. Is. The. Response. From. Here. And. The. Response. From. Here. Is. The. Call. And. The. Call. From. Here. Is. Complete. And.…

Fasrt mode in usage, even though I have it disabled

Where does the bug appear (feature/product)? Cursor IDE Describe the Bug In cursor Usage report it’s showing using “grok-4.7-high-fast” quiet often even though I have “Fast” mode disabled in Cursor App. It should only show “grok-4.7-high”. This means it’s drawing enormous amount of usage even on Ultra Plan that I am on. And yes I am 100% positive that I really never have or let the “Fast” mode enabled. Thanks Steps to Reproduce In Cursor App, set agent to Grok 4.7 High 500K Operating System…

Cursor Grok 4.7 wiped all data on secondary drive

Where does the bug appear (feature/product)? Cursor IDE Describe the Bug Grok 4.7 in agent mode caused a wipe in my secondary drive. A quoting bug caused rmdir to run against the drive root. Steps to Reproduce Don’t have any idea Expected Behavior It should not completely wipe all my data on disk drive Screenshots / Screen Recordings Operating System Windows 10/11 Version Information Version: 3.17.8 (user setup) VS Code Extension API: 1.128.0 Commit: 2fdd31c9f33f7fbe501f2d57772dc5bf64b63620…