Start with Standard when its query and source capacity cover a maintained collection. Premium becomes relevant when the larger workload, automatic website sync or branding requirement is part of the service you will actually run.
CustomGPT.ai’s plan choice combines knowledge capacity with the operation of the service. Queries describe use, source allowances describe the collection, seats describe the people managing the workspace and synchronization describes how information stays current. None of those numbers alone establishes whether the business will have a useful assistant.
This guide compares the official Standard and Premium snapshot reviewed on 14 October 2026. It then shows how to translate a hypothetical business requirement into a subscription decision. It is documentation-based buying guidance, not a claim of hands-on product performance or a guaranteed cost for an individual deployment.
Compare the current subscription references
Standard is listed at US$99 month to month, with an advertised US$89 monthly equivalent on annual billing. Premium is US$499 month to month, with an annual monthly equivalent of US$449. Both show a seven-day trial in this snapshot. Confirm the billing schedule and trial conditions in the actual checkout.
| Current allowance | Standard | Premium |
|---|---|---|
| Agents | 10 | 25 |
| Monthly queries | 1,000 | 5,000 |
| Documents per agent | 5,000 | 20,000 |
| Word storage | 60 million | 300 million |
| Staff seats | 3 | 5 |
Standard lists RAG API access. Premium adds automatic website synchronization and branding removal among its published differences. The table is a starting point for the review date; it is not an independent assessment of response quality. A larger plan does not repair outdated or contradictory source information.
Define what counts as one service
Write down the audience, source collection and expected route to the assistant. A website support agent, an internal procedure assistant and an embedded knowledge feature may require different deployments. Confirm how the vendor measures queries for that route and which integrations the selected plan supports.
For a hypothetical training business, the first service answers questions about a published course handbook. A second assistant could help staff find internal delivery procedures. Those are different audiences and source sets. The decision to separate them should follow the information boundary and user experience, not an urge to fill every available agent slot.
Count agents that have a defined owner and purpose. Include staging or testing requirements if the product’s current workspace model makes those relevant. A capacity of ten agents is useful room to organize work; it is not a recommendation to create ten assistants before one has a successful workflow.
Estimate queries from real questions
Use the current query definition and a representative sample of the intended experience. Someone asking a single question and someone refining an answer through several turns can create different usage patterns. Do not equate website page views, staff headcount or documents uploaded with monthly queries.
Estimate quiet and busy periods separately. A course launch may produce a short burst of questions even when normal traffic is modest. Include staff testing and quality review in the operating plan. Then compare the estimate with actual usage after a small release rather than treating a speculative forecast as a fixed number.
In the hypothetical training business, the owner records the volume of recurring questions during one course intake. That gives a more useful subscription estimate than multiplying every enrolled learner by an arbitrary number of daily prompts. The method is to observe a meaningful workload, not to manufacture a favorable return-on-investment claim.
Ordinary chat uses one query credit per message; enabled agentic actions can add costs. Plan & Act has a separate cost structure. Budget the functions the assistant actually uses rather than translating monthly queries into an equal number of complete conversations.
Clean the source collection before sizing it
The published document and word-storage allowances are substantial, but storage should follow the approved knowledge collection. Remove duplicate exports, old versions and irrelevant material before deciding which tier the business needs. A smaller collection with clear effective dates is often easier to maintain and review.
Check how the current source workflow counts documents and handles the formats you intend to use. A single large file can be awkward to update even if it fits the limit; a well-organized set of topic pages can make ownership clearer. Choose the structure that supports your content process rather than optimizing only for the largest numerical allowance.
Preserve the authoritative files outside the assistant workspace. The business should know which system owns the information and how a correction reaches the assistant. This makes plan changes and future migration easier to manage without losing the approved content.
| Knowledge budget | Standard | Premium | What to measure |
|---|---|---|---|
| Documents kept per agent | 5,000 | 20,000 | The approved collection already held. |
| Documents indexed monthly | 10,000 | 40,000 | New and refreshed material processed across the service. |
| Words processed monthly | 120 million | 600 million | Ingestion work, including repeated updates. |
Make update frequency a buying requirement
Automatic website synchronization is a meaningful Premium difference when the source website changes often. The official auto-sync documentation describes daily, weekly, monthly and custom schedules for eligible plans. Set the interval around the consequence of stale information and the rate at which the source changes.
For a handbook revised once per course intake, a deliberate source review may be sufficient. For a frequently updated public help site, scheduled sync can reduce repeated maintenance work. Confirm which source types the feature covers and verify a changed page reaches the assistant as expected during your own implementation.
Automatic synchronization is not an editorial approval process. If the underlying website publishes an error, a fresh sync can make that error current more quickly. Keep responsibility for source accuracy with the document or website owner.
Consider seats, branding and API use together
Three Standard seats or five Premium seats need to fit the people who will administer, maintain and review the service. Check the current roles and access controls for your deployment. Do not infer a permission model from the seat count alone, especially when internal documents are involved.
Branding removal can matter in a client-facing experience, but compare its value with the rest of the service. A discreet vendor label may be acceptable for one project, while another has a defined brand requirement. Treat the requirement as a product decision instead of assuming every business must buy the highest tier.
If using the RAG API, include implementation, connected systems and maintenance in the project budget. API availability establishes a route to integration; it does not supply the complete application, its permissions or its operational support.
Choose Standard when the current capacity and manual source process fit the first service. Choose Premium when larger usage, more source room, scheduled website updates or branding removal solve a specific requirement. Revisit the choice with usage and maintenance evidence before committing to a longer billing period.
Costs alongside this guide
Recorded subscription references. Check the selected billing term, current allowances, additional usage and tax.
Plans for this workflow
USD · US / global observations
CustomGPT.ai · Standard — monthly
- Initial term & commitment
- Initial term amountUS$99.001 month termPublished base monthly subscription rate; additional usage and tax are separate.
- Renewal
Check local tax at checkout
Price observed 2026-10-14
Plan details & sources
A focused business knowledge or documentation assistant.
10 agents; 1,000 queries/month; 5,000 documents/agent; 3 team members.
Choose another option if: Check update requirements and the query volume before expanding use.
Global USD pricing page; checkout country and tax not selected · Check local tax at checkout
Renewal: Future renewal amount and any additional usage should be checked before subscribing.
Check local tax at checkout
Prices checked 2026-10-14Official price source ↗Observed 2026-10-14
Official price evidence ↗CustomGPT.ai · Premium — monthly
- Initial term & commitment
- Initial term amountUS$499.001 month termPublished base monthly subscription rate; additional usage and tax are separate.
- Renewal
Check local tax at checkout
Price observed 2026-10-14
Plan details & sources
A broader knowledge workflow that needs source synchronization.
25 agents; 5,000 queries/month; 20,000 documents/agent; 5 members; sync and branding removal.
Choose another option if: The subscription does not replace source ownership and answer review.
Global USD pricing page; checkout country and tax not selected · Check local tax at checkout
Renewal: Future renewal amount and any additional usage should be checked before subscribing.
Check local tax at checkout
Prices checked 2026-10-14Official price source ↗Observed 2026-10-14
Official price evidence ↗Recorded prices can change. Check eligibility, taxes and the current full bill with the provider. Different currencies and services are shown separately.
Sources and verification
Editorial review 2026-10-19. Source check dates are recorded separately. Notes describe what each source supports and its limits.
View sources
- CustomGPT.ai current plansChecked Standard and Premium prices, queries, agent counts, source capacity, seats, branding and trial.
Limits: Official documentation reviewed on the stated date. Plan boundaries can change; this is not a hands-on product test. - CustomGPT.ai enterprise knowledge searchChecked Business knowledge search using the organization's sources.
Limits: Official documentation reviewed on the stated date. Plan boundaries can change; this is not a hands-on product test. - CustomGPT.ai agent settingsChecked Citation, response and source-grounding settings for an agent.
Limits: Official documentation reviewed on the stated date. Plan boundaries can change; this is not a hands-on product test. - CustomGPT.ai source auto-syncChecked Premium and higher plans support scheduled website synchronization with daily, weekly, monthly or custom schedules.
Limits: Official documentation reviewed on the stated date. Plan boundaries can change; this is not a hands-on product test. - CustomGPT.ai action costChecked Ordinary chat versus action query costs; no claim that one answer always costs one credit.
Limits: Official documentation checked on 18 October 2026; vendor documentation is not an independent performance test. No checkout, account entitlement or live product test was performed. - CustomGPT.ai usage limitsChecked Retained knowledge versus monthly indexing and processed-word budgets.
Limits: Official public documentation rechecked 19 October 2026. Describes documented features, not account entitlement, a final checkout, hands-on performance or conversion results.