Every leadership team I speak to in Melbourne has the same two questions about OpenAI: which ChatGPT plan do we actually put people on, and is it safe to feed it our data? The honest answers changed again in 2026, so here is where things stand.
ChatGPT is OpenAI’s chat product for getting work done with GPT models, and for most Australian businesses the right starting plan is ChatGPT Business at 20 USD per seat, because it gives you admin control and a contractual no-training guarantee without an enterprise-sized commitment. Everything else in this piece is the detail behind that sentence.
OpenAI, ChatGPT, and the API: what is what?
People use “openai” and “chatgpt ai” interchangeably, but for a buyer the distinction matters.
OpenAI is the company. ChatGPT is the product your staff log into. The API is the platform your developers build on. Get those three straight and most of the confusion disappears.
- ChatGPT is the assistant at chatgpt.com and in the desktop and mobile apps. You buy it per seat. This is what 90 percent of your people mean when they say they “use AI”.
- The OpenAI API is the developer platform. You do not log in and chat; your software calls it and pays per token of text in and out. This is how you embed GPT into your own products, internal tools, and automations.
- OpenAI is the vendor behind both, and the one whose data policies and pricing you are actually signing up to.
Most organisations end up needing both surfaces. ChatGPT handles the day-to-day knowledge work, and the API handles anything that has to live inside your own systems.
What is the GPT model line-up in 2026?
The current flagship family is GPT-5.6, released to general availability on 9 July 2026, and it comes in three tiers so you can trade intelligence against cost. OpenAI has moved to named tiers rather than a single monolithic model, which is a genuine improvement for cost control.
- GPT-5.6 Sol is the flagship, aimed at the hardest coding, analysis, and knowledge work. It is the one you reach for when quality matters more than price (OpenAI’s GPT-5.6 announcement).
- GPT-5.6 Terra is the balanced everyday model, with performance close to the previous GPT-5.5 at roughly half the cost.
- GPT-5.6 Luna is the fast, low-cost option for high-volume, simpler tasks.
All three ship with a one-million-token context window, which is large enough to drop a full contract set, a codebase, or a quarter of board papers into a single prompt. The earlier GPT-5.4 and GPT-5.5 models are still around, but for anything new I would default to the 5.6 family.
The practical takeaway: do not run everything on the flagship. Route simple, high-volume jobs to Luna or Terra and save Sol for the work that genuinely needs it. That routing discipline is where AI budgets are won or lost, a theme I cover in the best AI tools for Australian businesses in 2026.
How much does ChatGPT cost? The 2026 plans
ChatGPT spans six tiers from Free to Enterprise, and the jump that matters for any business is from the consumer plans (Free, Go, Plus, Pro) to the business plans (Business, Enterprise), because that is where the data-handling terms change.
Here is the current line-up. Prices are in USD per user per month.
| Plan | Price | Who it is for | Key differences |
|---|---|---|---|
| Free | 0 | Individuals trying it out | Core model access, limits on advanced features. Consumer data terms apply. |
| Go | 8 | Light personal users | Higher limits than Free, still a personal plan. |
| Plus | 20 | Power users, individuals | Advanced Voice, Deep Research, Agent Mode, custom GPT creation. Personal account. |
| Pro | 100 and 200 | Heavy individual users | Highest usage ceilings and priority access to top models. |
| Business | 20 (annual) / 25 (monthly) | Small to mid-size teams | Shared workspace, admin console, SSO, no training on your data. Two-seat minimum. |
| Enterprise | ~45 to 75 (negotiated) | Larger organisations | Everything in Business plus SCIM, audit logs, data residency, SLA. Larger seat commitment. |
Two things to flag. First, ChatGPT Team was folded into ChatGPT Business in 2026, so if your team is still on an old “Team” subscription you are effectively on what is now called Business. Second, the Business price dropped to 20 USD per seat on annual billing (25 USD month to month) in the April 2026 change, which narrowed the gap with Plus and made it much easier to justify moving staff off personal accounts (ChatGPT plan details).
ChatGPT Business vs Enterprise: which tier?
Choose Business for most teams and reach for Enterprise only when a specific control forces you to, such as Australian data residency, directory-based provisioning, audit logs, or a signed service level agreement.
ChatGPT Business gives you the essentials that make a plan safe to roll out: a shared workspace, an admin console, single sign-on, and the no-training data commitment. For a team of five to a couple of hundred, that is usually enough.
ChatGPT Enterprise adds the controls that larger and more regulated organisations need:
- SCIM provisioning so accounts are created and de-provisioned from your identity directory automatically.
- Role-based access control and audit logs for compliance evidence.
- Data residency, including the Australia region, covered in detail below.
- A negotiated SLA and longer data-retention controls.
Enterprise pricing is negotiated rather than listed, and reported 2026 contracts land somewhere around 45 to 75 USD per user per month depending on seat count and term. My advice to most mid-market clients: start on Business, prove the value, and only take on the Enterprise commitment once you have a named control you cannot meet on Business. If you are weighing this against the Microsoft stack, my Microsoft Copilot for business breakdown is worth reading alongside this.
Is ChatGPT safe for company data?
On the business tiers, yes, with the important caveat that the tier you are on decides the answer. OpenAI does not train its models on your inputs or outputs by default across ChatGPT Business, ChatGPT Enterprise, and the API. This is the single most misunderstood point in the whole conversation.
The reason personal accounts feel risky is that consumer plans (Free, Plus) have different, more permissive data settings than the business tiers. The moment you move staff onto ChatGPT Business, the terms change: your data is excluded from training by default, it is encrypted at rest and in transit, and OpenAI backs the arrangement with SOC 2 Type 2 and the ISO 27001 family of certifications (OpenAI business data commitments).
So the real data-safety risk in most organisations is not OpenAI. It is shadow usage: staff pasting sensitive material into personal Free accounts because you never gave them a sanctioned one. Rolling out ChatGPT Business is as much a governance control as it is a productivity tool. For the deeper enterprise position, OpenAI documents it under enterprise privacy.
Australian data considerations
If your data classification or a regulator requires content to stay onshore, OpenAI now supports Australia as a data residency region on ChatGPT Enterprise, ChatGPT Edu, and the API, so in-scope content can be stored at rest in-country. That was not true a year or two ago, and it changes the compliance conversation for government-adjacent and regulated clients.
Read the fine print, though. Residency covers data at rest: conversations, uploaded files, custom GPTs, and generated artifacts. Live inference still defaults to the United States (OpenAI on expanding data residency). For most commercial workloads that distinction is manageable, but if you handle data with strict cross-border rules you need to classify it and decide what is allowed to be processed, not just stored, offshore.
Two more Australian-specific notes. Data residency applies to new Enterprise workspaces rather than being retro-fitted, so plan the workspace setup deliberately. And residency is not a substitute for your own obligations under the Privacy Act and any sector rules; it is a control that helps you meet them.
Custom GPTs and agents: where the leverage is
The real business value is not chatting; it is building custom GPTs and agents that encode your process, your documents, and your tone, so staff get consistent output instead of ad-hoc prompting.
A custom GPT is a configured version of ChatGPT with its own instructions, knowledge files, and connected tools. Practical examples I have shipped or seen work:
- A proposal drafter loaded with your templates, past winning bids, and style guide.
- A policy assistant that answers staff questions grounded only in your current HR and IT policies.
- A first-line support GPT that drafts replies in your brand voice and knows your product documentation.
Beyond static GPTs, ChatGPT’s Agent Mode and the API’s agent tooling let you build assistants that take actions: pulling data, running research, and completing multi-step tasks with connectors into your systems. This is the same agentic direction the whole field is moving in, and if you are weighing OpenAI against the alternatives for agent work specifically, I compare them directly in Claude vs ChatGPT vs Gemini agents.
When to reach for the API instead of ChatGPT
Use ChatGPT for people and the API for products. If a human is going to sit and interact, buy them a ChatGPT seat. If GPT needs to run inside your own software, automatically and at scale, that is an API job.
The API is billed per token, and the 2026 rates reflect the same tiered thinking as the models. Per million tokens (input / output), GPT-5.6 Sol runs around 5 / 30 USD, Terra around 2.50 / 15 USD, and Luna around 1 / 6 USD, with cached prompt prefixes billed at a fraction of the standard input rate (OpenAI API pricing). For a high-volume automation, choosing Luna over Sol and using prompt caching can cut your bill by an order of magnitude with no meaningful quality loss on the right task.
The trap I see: teams building on the API when a custom GPT would have done the job in an afternoon, or buying ChatGPT seats when the real need was a single automated pipeline. Match the surface to the shape of the work, not to whichever one you heard about first.
The honest bottom line
- Put staff on ChatGPT Business (20 USD per seat) as your default. It fixes the shadow-usage risk and gives you the no-training data terms.
- Move to Enterprise only for a named control: Australian data residency, SCIM, audit logs, or an SLA.
- Default new work to the GPT-5.6 family, and route by tier (Sol, Terra, Luna) to control cost.
- Treat data residency carefully: onshore at rest is available, inference is not.
- The leverage is in custom GPTs and agents, not in more chatting.
OpenAI is not the only serious option, and I would never tell a client to standardise on one vendor without looking at the field. Google Gemini for business is worth a look if you live in Workspace, and the broader landscape is in the best AI tools for 2026. But if you asked me where a typical Australian business should start with OpenAI today, the answer is short: ChatGPT Business, a handful of custom GPTs, and a clear data policy.
I help Australian leadership teams pick, roll out, and govern tools like ChatGPT without the hype. If that is where you are, get in touch.