Grok for Professionals (2026): Plans, Limits, Fit

This page separates the model lineup and prices we verified from the questions you still need to resolve before buying. A listed usage allowance is not evidence of accuracy, suitability for your profession or confidentiality protections.

What is Grok, and which models does it offer now?

Grok is xAI’s model family. The verified lineup is Grok 4.7, Grok 4.7 Fast and Grok 4.6.

• Grok 4.7 is the flagship, released 2026-09-21. Its API id is grok-4.7. The recorded specifications are a 500k context, a knowledge cutoff of May 2026, and API pricing of $2 / $6 per 1M tokens (<200k prompt).
• Grok 4.7 Fast was released 2026-09-21. It is the same model with faster serving, available only in Cursor and Grok Build, not on the public API.
• Grok 4.6 is the previous model, released 2026-08-12.

Older versions still visible on vendor pages include Grok 4.5, Grok 4.20, Grok 4.1 Fast, Grok 4 and Grok 3. Do not mistake those references for the current lineup.

Keep API pricing separate from the subscription prices below. If you are buying a subscription rather than planning an integration, first confirm which model the plan includes. Our verified facts do not establish model access for each plan.

Which Grok plan fits your work and budget?

Free is the cautious starting point for an individual evaluating non-confidential work. Consider a paid plan only after checking its current access and usage terms against the tasks you actually need to complete.

Prices checked 2026-10-19 on the official pricing page.

These are USD list prices shown to a US visitor, excluding taxes and fees. For readers in the UK, Australia and India, local prices, currency and tax differ; check the vendor’s pricing page for your country.

For occasional use, try Free with public documents or invented examples. An event planner might assess a draft schedule; an engineer might review an explanation of a familiar concept. These are suggested evaluation tasks, not capabilities we have tested.

Plus is positioned by the vendor for social creators with limited AI usage. That description alone does not establish whether it suits an accountant, lawyer or healthcare professional.

Premium, Super and Ultra have verified usage multiples relative to Plus. Those figures do not tell us the practical allowance for your workflow or establish differences in answer quality. Before upgrading, identify what is limiting your work and confirm that the proposed plan addresses it.

For a firm, Teams and Enterprise are plans to investigate, not proof of business protections. Check both the commercial terms and the data terms before making a purchasing decision.

How did Grok do in our tests?

Grok is not in our test round yet. We have run no prompt or Claude Skill on it, so we have no score, ranking or result to report.

Our testing uses an API harness rather than the consumer app; see our methodology. That distinction matters when interpreting results for tested models, but there is no Grok result here to extrapolate to a subscription.

You can still assess whether it helps with your own work. Choose representative, non-confidential tasks whose answers you can check. Compare the output with your usual standard: does it preserve essential details, follow instructions and reduce editing time?

Do not judge usefulness from fluent wording alone. Record what needed correction and whether checking the answer took longer than doing the task yourself. Your trial can inform a purchase without becoming evidence that Grok is reliable across your profession.

What should you check before pasting confidential data?

We have not verified the vendor’s data terms for this page. We cannot confirm retention periods, whether submitted data is used for training, or what business protections apply.

Before pasting client, patient or employee data, read the specific plan’s data and privacy terms and your professional confidentiality rules. If you work within an organization, also follow its approval process.

Do not infer protections from the names Teams or Enterprise. The verified facts contain no data-protection commitment for either plan.

For an initial trial, use public material or fictional examples rather than confidential records. Removing names alone should not be treated as permission to upload a document: other details may still identify the people or business involved. If you cannot establish that submission is permitted, keep the material out.

This caution applies beyond patient notes and legal files. Consider whether budgets, supplier quotes, employee records, property documents or contracts contain information you are obliged to protect.

Confidentiality duties also depend on your profession: our pages for lawyers, healthcare professionals and accountants list the rules that apply in your country.

When should you avoid relying on Grok?

Do not treat Grok as the final authority for consequential professional decisions. Tax, legal, health and property work must be checked by a licensed professional in your jurisdiction.

There are several practical reasons to stop and verify:

• Confidentiality is unresolved. A paid plan does not resolve the missing evidence about data handling.
• The answer depends on current information. Grok 4.7 has a knowledge cutoff of May 2026. Check current authoritative sources for changes to rules, prices or other time-sensitive information.
• The calculation has consequences. Independently verify tax computations, engineering calculations and dosage figures rather than accepting a plausible-looking answer.
• You cannot evaluate the output. If you lack the expertise or source material needed to check it, do not turn the response directly into client-facing advice.

The buying question is not simply whether Grok can produce a draft. It is whether you can review that draft safely and efficiently within your professional obligations.

For how the models did on real professional tasks, read our test-round report; ready-made task instructions for Claude are in the Claude Skills library.

FAQ