ChatGPT 6.1 Terra: What GPT-6.1 Terra Is and Who Can Use It

“ChatGPT 6.1 Terra” refers to GPT-6.1 Terra, a new OpenAI model announced in late October 2026 that is built to be fast and cheap while handling computer use (reading the screen and acting across apps), visual reasoning and everyday writing. It is available to paid ChatGPT plans in ChatGPT Work and Codex, and to developers through the API as gpt-6.1-terra, but not on the Free plan.

Below is what the model actually does, who can use it, what it costs, and how it compares with OpenAI’s other current models. Everything here reflects launch-week information as of late October 2026, and details may change as the rollout continues.

What GPT-6.1 Terra is

OpenAI describes GPT-6.1 Terra as a “fast, efficient model” that pairs sharper visual reasoning and precise computer use with clearer, more natural writing, according to its launch announcement. The pitch is a model for “the whole workday”: examples OpenAI gives include decoding technical drawings, drafting proposals, filling forms and editing videos.

Terra sits in the middle of OpenAI’s current naming scheme. In the Hacker News launch thread, an OpenAI employee placed the models in this order of capability:

• Luna (smallest)
• Terra
• Sol
• Astra (most capable)

So Terra is not the flagship. It is the efficient workhorse, meant to cover more everyday tasks for the same plan allowance or API budget.

A note on the release date

Coverage disagrees by a day. 9to5Mac dates the launch to October 26, 2026, while Developers Digest gives October 27. The difference is most likely time zones. Either way, it is brand new.

Who can use it

Here is how access breaks down at launch, based on OpenAI’s announcement and launch coverage.

A few points are less certain:

• Business plans. One launch write-up adds Business to the eligible list, but OpenAI’s announcement names Plus, Pro, Enterprise and Edu. If you are on Business, check with your workspace admin.
• Codex CLI version. One source reports that the Codex CLI needs version 0.159.0 or later to select Terra. We could not confirm this in OpenAI’s own material, so update to the latest official release if the model doesn’t show up.
• Where to click. We found no official step-by-step for selecting Terra in the ChatGPT interface, so we won’t guess at menu names.

How to try it

1. Confirm you are on an eligible paid plan (Plus, Pro, Enterprise or Edu).
2. Open ChatGPT Work or Codex, not a regular Chat conversation.
3. Look for GPT-6.1 Terra in the model options there. Rollouts are often gradual, so it may take time to appear.
4. For Codex on the command line, update the CLI through its official channel before trying to select the model.
5. Developers can switch the model parameter in API calls to gpt-6.1-terra.

What it costs

In ChatGPT, OpenAI says your plan “goes further” with Terra: usage allowances cover more tasks than with GPT-5.6 Terra or the Sol and Astra models. OpenAI has not framed this as a separate price; it is about how much work fits into your existing plan.

For the API, OpenAI says standard prices are 50% lower than those of GPT-5.6 Terra and GPT-6.1 Sol. The reported per-token numbers are:

Cached input for Terra is reported at $0.05 per million tokens. One source also reports Batch and Flex processing at 50% less and a Fast mode at 2x the standard rate. Confirm these on OpenAI’s official pricing page before you build a budget around them.

Context window: unconfirmed

Data Studios reports a context window of roughly 1,050,000 tokens, but another source says no limit was published. Treat the 1M figure as unconfirmed for now.

How well it performs

All of these benchmark figures come from OpenAI unless noted, so they are vendor claims, not independent tests.

• Agents’ Last Exam (medium reasoning effort): about 15 percentage points higher than GPT-5.6 Terra at roughly one-third the cost.
• Versus GPT-6 Astra: OpenAI says Terra matches GPT-6 Astra on this agent benchmark at one-tenth the price per token.
• OSWorld 2.1: reportedly beats GPT-5.6 Terra’s best score. OpenAI did not publish a direct Sol comparison here.
• Terminal-Bench 4.0 (max effort): 50.3%, versus 21.5% for GPT-5.6 Terra, per DataCamp’s roundup.

Independent numbers are more mixed. DataCamp cites Artificial Analysis giving Terra 48 at max effort on its overall index, against 52 for Sol. That supports the general picture: Terra is close to Sol on many tasks, but not equal.

Terra vs. Sol vs. Astra: which to pick

OpenAI is direct about Terra’s limits: “For the most complex coding tasks, GPT-6.1 Sol and GPT-6 Astra remain the better choice.” Launch coverage and developer discussion suggest a rough split:

In the Hacker News thread, the OpenAI employee suggested Astra if you care only about quality and Sol if you’re somewhat price-sensitive. Commenters also said Terra may be better at technical communication and similar on vision and computer use, while being much faster. Those are early impressions, not measurements.

A practical approach:

1. Start with Terra for routine screen, document and writing tasks.
2. Switch to Sol for complex coding or multi-step agent work.
3. Move to Astra only when Sol fails on your own test cases.

The bottom line

GPT-6.1 Terra is OpenAI’s efficiency play: a mid-tier model that handles on-screen actions, visual inputs and writing well enough for most everyday work, at half the API price of Sol. If you’re on Plus, Pro, Enterprise or Edu, look for it in ChatGPT Work and Codex. If your work is mostly hard coding, OpenAI’s own advice is to stay on Sol or Astra.

Availability, pricing and benchmark details are from launch week (late October 2026) and may change, so check OpenAI’s announcement and pricing pages for the latest.