Which ChatGPT Model Should You Use? OpenAI Now Has Four

Which ChatGPT model should you use? For most SMEs the answer is Terra or Sol, not the most expensive tier, now that OpenAI have released GPT-6.1 Terra on 26th October, 2026.

If your team already pay for ChatGPT, the tier names Luna, Terra, Sol and Astra probably mean little. We explain what each does, what it costs and how to choose between them.

This guide reflects OpenAI's help centre and press coverage on 27 October 2026. Model names and limits change often. Check your own plan before you act.

What Changed on 26 October: A Cheaper Everyday Model

OpenAI released GPT-6.1 Terra on Monday 26 October 2026, according to Bleeping Computer. It replaces GPT-5.6 Terra. The Next Web explains that it can view a screen, understand it and act by clicking, typing and moving between apps.

OpenAI's examples include entering invoices into accounting software, filling in forms and drafting proposals. Plus, Pro, Enterprise and Edu users can reach it in ChatGPT Work and Codex. Developers can call it through the API.

The API price is $1 per million input tokens and $5 per million output tokens, half the price of the model it replaces. Independent testers at Artificial Analysis found that the cost of a task at maximum effort fell from $1.40 to $0.40.

Most other headline claims come from OpenAI's own tests. OpenAI say Terra matches the top-tier Astra model on one benchmark at a tenth of the price. The Next Web notes that those figures are not independently verified.

Which ChatGPT Model Should You Use? The Four Tiers Explained

OpenAI sell four tiers. In order of price and capability they are Luna, Terra, Sol and Astra, according to 9to5Mac. We think of them as a trainee, a capable assistant, a senior specialist and an expert you call in rarely.

Luna is the fastest and lowest-cost tier, according to OpenAI's help centre. It is the default for Free and Go users. We would use it for simple repeat jobs such as sorting enquiries or tidying short text.

Terra offers a balance of capability, speed and cost for everyday work, in the words of the help centre. It is the screen-reading, form-filling tier. It does not appear in standard chat and lives in ChatGPT Work, Codex and the API.

OpenAI designed Sol for complex work across coding, knowledge work and research. Plus, Pro, Business and Enterprise plans include it. Free and Go plans do not.

Astra is the top tier. It powers GPT-6 Pro in chat for Pro, Business and Enterprise plans. Plus plans reach it in ChatGPT Work and Codex.

The help centre warns that a Pro subscription does not include unlimited use of GPT-6 Pro.

What Does Each Model Cost?

The API prices per million tokens, input then output, are $0.10 and $0.50 for Luna and $2 and $10 for Sol, according to OpenAI's launch page. Terra costs $1 and $5. One third-party report lists Astra at $10 and $50, which makes it five times the price of Sol.

Most staff pay a flat subscription and never see these prices. They matter if you build tools on the API or run agents that make thousands of calls. Our guide to AI agent costs for SMEs shows how quickly those bills grow.

Our View: Start With the Cheapest Model That Does the Job

For most SMEs, we would start with Terra for repeat computer tasks and Sol for thinking work. We would rarely reach for Astra. On Artificial Analysis's Intelligence Index at maximum effort, Terra scores only 4 points below Sol and 10 points above Luna, at half the price of Sol per token.

OpenAI still recommend Sol or Astra for the hardest coding, according to Bleeping Computer.

Take three steps. First, list your repeat tasks and sort them into simple, everyday and complex.

Second, run one task from each group on the matching tier, starting with the cheapest. Record the quality and the time saved.

Third, give one person, ideally your finance or operations lead, the job of reviewing the results each month.

An accountancy practice might test Terra on invoice entry in a copy of its software. A marketing agency might compare Sol and Terra on first drafts of monthly reports. A professional services firm might use Luna to sort incoming enquiries.

Test any tool that acts on a live system on a copy of your data first.

If you have no AI policy yet, start with our AI policy guide before you let any tool act inside your software. If you use ChatGPT but have not compared tools, you sit at the Curious stage of the AI confidence journey. Step one is enough for this month.

Our Claude vs ChatGPT comparison shows how ChatGPT sits against its main rival.

Why Not Just Pick the Top ChatGPT Model?

Choosing Astra for everything has a fair case because it removes the decision and handles the hardest work. It costs five times as much as Sol per token and carries usage limits on Pro plans, according to the help centre. A cheaper tier that does the job leaves budget and allowance for the work that needs Astra.

FAQs

When Did OpenAI Release GPT-6.1 Terra?

OpenAI released it on Monday 26 October 2026, according to Bleeping Computer and The Next Web. It replaces GPT-5.6 Terra.

Can I Use GPT-6.1 Terra in Standard ChatGPT Chat?

No, because the help centre says Terra is not selectable in standard ChatGPT conversations. Plus, Pro, Enterprise and Edu users can use it in ChatGPT Work and Codex. Developers can use it through the API.

Which ChatGPT Model Do Free Users Get?

Free and Go users get Luna by default, according to the help centre. Sol is not available on those plans.

Why Do the ChatGPT Model Names Keep Repeating?

OpenAI reuse the tier names across versions. Their documents mention GPT-5.6 Sol, GPT-6 Sol and GPT-6.1 Sol. Check which version your plan uses before you compare prices or limits.

Your Next Step After Choosing a ChatGPT Model

We are an independent Worcester-based AI consultancy for SMEs. We are not tied to software sales and have no stake in which supplier you pick. Sophie Delroy, Managing Director at Build Group, said our assessment's recommendations 'gave us clear next steps for our AI Workshop'.

Deciding which ChatGPT model you should use is easier once you know which of your tasks matter most. Complete our free AI Readiness Assessment and you receive a personalised report in two to three minutes.

The assessment leads to our AI Workshop, where we spend half a day to a full day with your leadership, operations and finance leads and map where AI saves time and money in your business. You then choose a tier on evidence from your own tasks and not on a vendor's benchmark.