GPT-6.1 Terra Pro vs GPT-6.1 Terra: Same Model, Two Different Doors

GPT-6.1 Terra Pro and GPT-6.1 Terra are the same model. That is the whole answer, and it is also where the useful part starts: the name people search for and the name OpenAI publishes point at one set of weights, one 1.05-million-token context window and one API identifier, gpt-6.1-terra, but they can lead to two very different bills. One door is a ChatGPT Pro subscription with an included allowance. The other is metered API access billed per token. Choosing between “Terra Pro” and Terra is really choosing between those two doors.

That is a more interesting decision than the naming question suggests, because the doors do not just change the price. They change the surface, the limits, what you can automate and what you cannot. And one of them has a deadline this week that affects how much Terra a Pro 200 subscriber actually gets.

The one-line answer

Same weights. Same 1.05-million-token context. Same $1 input and $5 output list price. Same knowledge cutoff of June 1, 2026. GPT-6.1 Terra Pro is not a bigger, faster or more capable sibling of GPT-6.1 Terra; it is GPT-6.1 Terra reached through a different commercial door. The model does not change when you cross the threshold. The bill, the surface and the limits do.

What the Pro door actually is

OpenAI’s help centre, updated October 27, lists three monthly Pro plans: Pro 100 at $100, Pro 200 at $200 and Pro 500 at $500, above Plus at $20. Terra is included for Plus, Pro, Enterprise and Edu users in ChatGPT Work and Codex. That is the precise sense in which a Pro subscriber has “Terra Pro”: they have Terra, inside the Pro plan, in the Work and Codex surfaces.

What the plan does not do is turn Terra into a different model.

• No separate model ID — the API identifier is gpt-6.1-terra on every plan.

• No separate benchmark — OpenAI published one set of Terra numbers, not a plan-specific set.

• No speed tier — Ultrafast is available only on Pro 500, and only for GPT-6 Astra and GPT-6.1 Sol. Terra has no Ultrafast option, so a higher Pro tier does not buy a faster Terra.

• No separate context window — 1,050,000 tokens, with 128,000 maximum output, regardless of plan.

The plan changes the allowance and the surface. OpenAI does not publish a token-equivalent for the allowance, which is the central opacity in this comparison: you can watch the meter move but not see the size of the tank.

The deadline worth knowing about

Pro 200 subscribers who were active at any point from September 22, 2026 through 10 a.m. Pacific on September 29, 2026 are grandfathered onto the older, larger allowance through October 29, 2026. After that date the price stays $200 and the included usage drops to the lower allowance offered to new subscribers. If you are planning a Terra-heavy Work or Codex workflow around a Pro 200 seat, the allowance you have this week is not the allowance you will have next week, and OpenAI says only affected users get an email.

The deadline does not change the model, the price per token or the context window. It changes how many tasks the same seat covers, which is the number that matters if Terra is doing your invoice queue or your document drafting. Anyone budgeting that work off this week’s throughput should re-check after October 29.

What the API door actually is

• Standard — $1 per million input tokens, $0.05 per million cached input, $5 per million output.

• Cache writes — $1.25 at 30-minute retention, $2.50 at 12-hour retention.

• Long context — above 272,000 input tokens, twice the input and cache rate and 1.5 times the output rate for the whole request.

• Batch and Flex — 50% of standard.

• Surface — no interface, no memory, no file uploads, no deep research. You build the application around the model.

There is no seat and no allowance. Every call is metered, which means every call is visible, attributable and automatable. For a product team that is the point; for a person who wants a workspace it is a tax on building things ChatGPT already does.

The break-even, worked out

Start with the only published per-task number. Artificial Analysis measures GPT-6.1 Terra at $0.40 per Intelligence Index task at max effort. On that mix, $100 of API spend buys about 250 tasks, $200 buys about 500 and $500 buys about 1,250. A Pro 100 seat, by contrast, is not sold in tasks and OpenAI publishes no conversion, so the comparison is directional rather than exact.

Now use a lighter workload. A document-extraction job at 2,000 input and 500 output tokens costs $0.0045 at Terra’s list price. $100 buys roughly 22,000 of those before caching. If your automation runs at that scale, the subscription is not the cheaper door; the API is dramatically cheaper, and the plan is buying the interface and the Work and Codex integration rather than the tokens.

The reverse is also true. If your usage is a person working inside ChatGPT for a few hours a day, with file uploads, memory and deep research, metering the same work would be awkward to instrument, and the seat buys a product the API does not include. The break-even is not a single number; it is a question about whether you are buying tokens or buying a workspace.

What each door gives you that the other does not

• The plan gives you ChatGPT Work and Codex interfaces, file uploads, memory, deep research, image creation and no key management, inside an allowance with published reset behaviour but unpublished size.

• The API gives you automation, Batch and Flex discounts, cache control, routing, observability and the ability to put Terra behind your own product, with no seat and no interface.

• The plan cannot be routed. A subscription allowance is tied to a person and a workspace; you cannot fail a ChatGPT Work request over to another provider, and you cannot compose Terra with another model inside one call.

• The API has no memory or workspace. Anything the plan gives you as a product, the API makes you build.

Why the name keeps spreading

If you meet “Terra Pro” in a price list, the useful assumption is that it means one of three things: Terra inside a Pro plan, a third-party label, or a confusion with the Astra Pro configuration. In none of those cases is there a second model to evaluate.

The check is the same one that settles any ambiguity. Read the model string on the response, confirm the 1.05-million-token context and the June 1, 2026 cutoff, confirm that none is accepted as a reasoning level, and confirm the $1 input and $5 output card. A provider that cannot produce those four facts is not serving Terra, whatever the listing calls it.

Where OrcaRouter fits

OrcaRouter is for the API door. One key reaches more than 200 models at provider list price with no markup, so a vendor price change is reflected on the same terms, and automatic failover and a routing DSL let you compose task classes across models rather than hard-coding one. Model fusion is there for work that benefits from several models answering together.

The honest limit: GPT-6.1 Terra is not in our catalogue as of October 27, so Terra calls go to OpenAI’s API today. GPT-6.1 Sol is in the catalogue at $2 input and $10 output. And no router can help with the subscription door at all, because a ChatGPT plan is not a routed endpoint. If your work is inside Work or Codex, the plan is the only door; if your work is programmatic, the API is, and we are one way to reach the models that are listed.

Which door to take

Take the plan if the work happens in ChatGPT, the value is in the interface and the integrations, and you can live with an allowance whose size OpenAI does not publish. Take the API if the work is automated, volume is the point, and you need the Batch discount, cache control or a route you can fail over. Do not pay for a “Pro” model that does not exist, and do not assume a Pro seat makes Terra faster: the only Ultrafast tiers OpenAI documents are on GPT-6 Astra and GPT-6.1 Sol. The model is the same on both sides of the door. The decision is about which side your work lives on.