# Choosing Dots or Grok Bot starts with your account

> Compare OpenAI Dots and Grok Bot by account eligibility, cloud and local computers, model control, permissions, and how included usage becomes extra charges.

- Published: 2026-09-30
- Last reviewed: 2026-09-30
- Data as of: 2026-09-30
- Article slug: dots-vs-grok-bot
- Author: [Glevd](https://x.com/glevd)
- Reading time: 8 minutes
- Topics: openai, dots, grok-bot, ai-agents, permissions, pricing
- Data policy: Dated analysis; static benchmark claims are retained.
- Canonical URL: https://benchlm.ai/blog/posts/dots-vs-grok-bot

OpenAI Dots and Grok Bot make different demands on your account and usage. Dots comes with eligible ChatGPT access. Grok Bot uses a Cursor account.

Your first Dot is included with eligible Pro or Business Premium. Grok Bot is included with paid individual Cursor plans, self-serve Teams, or qualifying linked individual subscriptions.

As of September 30, 2026, we'd start with the account you already qualify for, then check its allowance and computer permissions. Neither product has been tested here.

This comparison uses public documentation. We haven't onboarded a paid account, inspected an invoice, or run the same task in both products. The comparison helps choose a starting configuration. It can't establish which agent finishes more work correctly.

| Decision | OpenAI Dots | Grok Bot |
| --- | --- | --- |
| Which account qualifies? | Eligible ChatGPT Pro or Business Premium, subject to gradual rollout and region. Enterprise has an admin-enabled beta.<br><br>[Access](https://help.openai.com/en/articles/20001530-getting-started-with-your-dot) | Paid individual Cursor plans and self-serve Teams, or eligible individual SuperGrok or X Premium+ links. Enterprise enablement uses the account team.<br><br>[Access](https://cursor.com/help/grok-bot/plans) |
| What uses the allowance? | Conversations are included. Deeper work has an allowance, and delegated Codex or ChatGPT Work tasks consume their usual usage.<br><br>[Usage](https://openai.com/index/introducing-dots/) | Weekly included usage, followed by optional on-demand usage billed through Cursor. Linked grants do not add together.<br><br>[Billing](https://cursor.com/help/grok-bot/plans) |
| What computer am I granting? | Each Dot has a cloud computer. Local access is optional and initially off.<br><br>[Computer access](https://help.openai.com/en/articles/20001530-getting-started-with-your-dot) | One persistent cloud computer per user. All that user's Bots share its files, browser sessions, and logins.<br><br>[Shared computer](https://docs.x.ai/grok-bot/faq) |
| Where do I set action boundaries? | App permissions and Custom Rules, with required confirmations and handoffs retained.<br><br>[Safeguards](https://openai.com/index/how-we-build-safety-security-and-privacy-into-dots/) | Auto-review rules and action approvals. Local command execution has separate controls and asks by default.<br><br>[Controls](https://docs.x.ai/grok-bot/security) |
| Can I choose the model? | GPT-6 Astra is documented. A Dot model picker or model API-key option is undocumented in the reviewed sources.<br><br>[Launch](https://openai.com/index/introducing-dots/) | Cursor selects serving models. No customer model picker or fixed vendor set is guaranteed.<br><br>[Model selection](https://docs.x.ai/grok-bot/security) |
| What cost remains unknown? | The account's exact deeper-work allowance and future additional-Dot prices.<br><br>[Launch terms](https://openai.com/index/introducing-dots/) | Numeric Bot allowances and a complete per-task bill have not been established here.<br><br>[Plan terms](https://cursor.com/help/grok-bot/plans) |

## An eligible account gives you a starting point

Dots access depends on more than the name of a ChatGPT subscription. Pro users in the initial rollout must be outside the European Economic Area, Switzerland, and the UK. Business Premium can qualify in supported ChatGPT regions, while Enterprise's beta needs an administrator to enable it. Access is gradual, so an eligible plan doesn't promise that the feature is available to your account immediately. Dots is [unavailable to users under 18](https://help.openai.com/en/articles/20001529-dots-privacy-security-and-safety-faqs). [Dots access conditions](https://help.openai.com/en/articles/20001530-getting-started-with-your-dot).

If you already have qualifying ChatGPT access, testing one responsibility in Dots can help you decide whether you need another product. Moving to a higher Pro tier won't change Pro's region exclusions. Business Premium has different regional eligibility. For a workspace, check the actual Premium seat and administrator settings. “Our company pays for ChatGPT” doesn't settle those fields. The [Dots guide](/blog/posts/openai-dots-guide) covers current subscription and seat prices, the Business minimum purchase, and documented setup steps.

Grok Bot gives an existing Cursor user a different starting route. Its [billing page](https://cursor.com/help/grok-bot/plans) includes paid individual plans and self-serve Teams. Qualifying individual SuperGrok and X Premium+ subscriptions can also provide usage through an account link. You still sign in with a Cursor account. Check both accounts before linking: Cursor says a SuperGrok or X Premium+ link is permanent and cannot be unlinked or moved to another Cursor account. Its [linking guide](https://cursor.com/help/grok-bot/supergrok) directs a wrong-account case to support.

We'd inspect the eligible account before buying access elsewhere. That recommendation loses if the product doesn't support the connections your job needs, its allowance is insufficient, or its permission boundary is unacceptable. Grok's geographic eligibility isn't established in this comparison, so Dots' exclusion list cannot become a claim that Grok is available everywhere else.

## Included access still needs a usage check

Dots separates conversation from delegated work. Talking to a Dot doesn't consume ChatGPT usage limits. Having it start a Codex or ChatGPT Work task uses those products' normal allowance. OpenAI also includes an allowance for deeper Dot work without publishing a fixed task count in the launch terms. [Dots usage](https://openai.com/index/introducing-dots/).

Work and Codex share the plan's included usage. Task size, selected model, and reasoning settings affect how much is used, according to [OpenAI's usage guide](https://help.openai.com/en/articles/20001516-managing-usage-with-gpt-6-astra-in-work-and-codex). Asking for a proposed change and delegating the job that makes it therefore need different budget checks. Inspect usage after the job rather than counting messages and assuming they measure the work.

Grok Bot has a weekly grant and optional extra usage billed through Cursor. Once a task is running, it can finish past the on-demand monthly limit. Keep that behavior in mind before leaving a large job running. [On-demand terms](https://cursor.com/help/grok-bot/plans).

Paying for Cursor and SuperGrok together doesn't sum their Bot allowances. [Grok’s FAQ](https://docs.x.ai/grok-bot/faq) says the larger applicable allowance is used. Cursor's plan page says the grants don't stack. Selecting one allowance and adding two allowances are different operations, so those rules agree.

Cursor and SuperGrok can both keep billing while Grok Bot uses one allowance.

Before buying a second subscription to extend agent usage, check the current grant and what additional work would actually unlock. Neither product has a measured all-in job cost here, and an unknown allowance can't support a cheapest-product verdict.

## A second Grok Bot has the same computer access

Grok Bot's named roles all use the same persistent computer on a user account. Research and sales Bots share its browser sessions and files. Its [FAQ](https://docs.x.ai/grok-bot/faq) explicitly says separate Bots are not a security boundary.

Suppose you want one Bot for each client. Different names and instructions can help describe the jobs, but they don't separate the client logins already on that computer. If client credentials must be isolated, this account arrangement is the wrong basis for the workflow. Resolve that requirement before connecting the accounts, rather than assuming a second Bot creates a second machine.

Dots has its own cloud computer and can work without using your laptop. Optional local access begins switched off; work on a connected local device needs that device online with ChatGPT open. [Dots computer access](https://learn.chatgpt.com/docs/dots/getting-started). Cloud tasks and jobs using local files have different device requirements.

Each Dot's own computer doesn't establish separately scoped app connections, though. OpenAI says [plugin permissions are shared across Dots, ChatGPT, Work, and Codex](https://help.openai.com/en/articles/20001529-dots-privacy-security-and-safety-faqs). Review the existing grants as well as the new ones.

For actions, Dots has [Custom Rules and required confirmations](https://openai.com/index/how-we-build-safety-security-and-privacy-into-dots/). Grok has [Auto-review rules and separate local-execution controls](https://docs.x.ai/grok-bot/security), with command approval by default. These are controls to inspect, not evidence that either product will respect every boundary perfectly. Your request to draft a message should state whether sending is authorized, whichever account you choose.

## Cursor chooses models and Dots' picker is undocumented

Dots documents GPT-6 Astra as its underlying model. The reviewed launch and setup material doesn't document a customer model picker or a bring-your-own-model-key path for the Dot. Work and Codex have their own model settings. Those settings don't establish that you can change the Dot itself. [Dots model and delegated work](https://openai.com/index/introducing-dots/).

Grok Bot is more explicit about this limit. [Cursor chooses its serving models](https://docs.x.ai/grok-bot/security), the mix can change, and the product has no customer-facing picker. Grok chat or the xAI API can offer a model without making it selectable inside a Bot. Billing follows the model that served the request, which is another reason to inspect the account's usage rather than budget from the product name alone.

That matters if your buying decision starts with a model leaderboard. Our [agent benchmark results](/llm-agent-benchmarks) describe model and harness evaluations. They don't measure the full Dots or Grok Bot experience: connected accounts, permission checks, persistence, or recovery when the work goes wrong. Our [scoring methodology](/methodology) explains the scope of those results.

If a configurable model or a local-only runtime is a firm requirement, neither product is established as a fit by these sources. Our [broader personal-agent comparison](/ai-agents) records configurable alternatives and their own billing and hosting obligations. Model freedom has consequences beyond the picker, so check the inference route as well as the machine running the agent.

## Compare one result before paying for another plan

Our starting choice is Dots for an eligible ChatGPT user who wants a Dot to manage Work or Codex tasks and is willing to monitor their shared usage. That fit fails if Dots hasn't reached the account, the required connection is unavailable, or the work needs control over the Dot's model. Familiarity with ChatGPT doesn't settle those questions.

Grok Bot is a reasonable starting choice for an eligible Cursor or linked-subscription user who wants a hosted computer and accepts its account-wide Bot logins. It is the wrong arrangement when the job needs each Bot to have separately isolated credentials, or when the user needs to pin the serving model. This recommendation follows the documented configuration, not a task-completion test.

One useful first comparison could be a deadline check: read an approved project source, list the deadlines that changed, and link each change to its evidence. Require approval before editing shared files or messaging anyone. This is a suggested evaluation we haven't run. It produces an artifact you can compare with the source without giving the agent a broad, hard-to-inspect responsibility.

Inspect the output, the actions taken, any approval requests, and the usage after that job. If the existing product misses a requirement, use the same source material and success criteria when evaluating the alternative. Don't widen its permissions just to make the first result look better. Justify a plan change by naming the unmet requirement or observed result that the other product must improve.

## Questions

### Can I link SuperGrok without buying a Cursor plan?

Yes. Cursor's billing page says an eligible individual SuperGrok, SuperGrok Plus, SuperGrok Heavy, or X Premium+ link can grant Grok Bot usage without a paid Cursor plan. You still use a Cursor account. Linking does not create a Cursor subscription, and any existing subscription continues under its own billing terms.

### Can I choose Grok's latest model inside Grok Bot?

Grok Bot has no customer-facing model picker in the security documentation checked September 30, 2026. Cursor selects the serving models, and the mix can change. A model's availability in Grok chat or the xAI API therefore does not establish that you can select it for a Grok Bot task.

### Can local computer access keep all task data on my machine?

Local access lets the agent work with your device; it does not make the entire product local. Dots has its own cloud computer, and Grok Bot runs on Cursor-hosted computers. Neither reviewed source set establishes a self-hosted, local-only version. Review the connected apps and model processing before giving either product sensitive files.