GPT-6.1 Terra: does it really cost $2 per million tokens?

On Monday, October 26, at 2:06 p.m. ET, OpenAI announced GPT-6.1 Terra, its mid-tier model built for computer use, visual reasoning and writing. Within minutes a price was making the rounds on X: $2 per million input tokens and $10 per million output tokens.

That pair is on OpenAI's pricing page, but it is the Fast mode price. Standard pricing is half of that. I went through the announcement, the API model page and the pricing page to sort what is confirmed from what OpenAI claims and what is just rumor.

The price, per OpenAI's own pages

Standard API pricing per million tokens, from the model page and the pricing page:

• Input: $1.
• Cached input: $0.05, which is 5% of the regular input rate.
• Cache writes: $1.25 with 30-minute retention, $2.50 with 12-hour retention.
• Output: $5.

The variants, from the same model page:

• Batch and Flex: 50% below standard, so $0.50 input and $2.50 output.
• Fast mode: twice standard, so $2 input and $10 output. The pricing page notes that what used to be priority processing was renamed Fast mode on July 30, 2026.
• Prompts over 272K input tokens: twice the input and cache rates and 1.5 times the output rate, for the whole request. At standard, that is $2 input and $7.50 output.
• Regional processing (data residency): a 10% premium where available.

Where "$2/$10" comes from

The post I saw reshared most lists "GPT-6.1 Terra: $2/$10", "GPT-6.1 Sol: $4/$20" and "GPT-6 Astra: $20/$100". All three pairs match the Fast table on the pricing page. The post even says so in its last line ("at Fast speed"), but what gets copied is the number at the top.

There is a second coincidence that makes it worse: $2 input and $10 output is also GPT-6.1 Sol's standard price. Read the wrong row and you get the same pair.

A generic example to make the gap concrete: a request with 10,000 input tokens and 1,000 output tokens, no cache, costs $0.015 at standard and $0.03 in Fast mode. That is my own math from the published prices. At volume, leaving Fast on when you do not need it doubles the bill.

"50% cheaper": right, with a caveat

The announcement says Terra's standard prices are 50% lower than GPT-5.6 Terra's and GPT-6.1 Sol's. Against Sol, the math is exact: $2/$10 versus $1/$5. Against GPT-5.6 Terra, which the pricing page lists at $2 input and $12 output, input drops 50% and output drops a bit more, roughly 58% by my math.

What Terra ships with

From the model page:

• A 1,050,000-token context window, with up to 922,000 input tokens and 128,000 output tokens.
• Knowledge cutoff: June 1, 2026.
• Text and image in, text out. No audio or video.
• Configurable reasoning.effort: none, low, medium (the default), high, xhigh and max.
• The computer_use tool in the Responses API, alongside web search, file search, hosted shell, MCP and others. Tool calling requires the Responses API.
• No fine-tuning.

Per the announcement, it is available to Plus, Pro, Enterprise and Edu users in ChatGPT Work and Codex, and in the API as gpt-6.1-terra.

One practical point about that 1.05M window: once a request goes past 272K input tokens, the whole request is billed at the long-context rate. Fitting in the window and being worth filling are two separate questions.

Proven, claimed, rumored

Confirmed on official pages: standard, Batch, Flex and Fast pricing; the surcharge above 272K tokens; context window, input and output limits, knowledge cutoff, modalities and tools; availability in ChatGPT Work, Codex and the API.

According to OpenAI, not independently verified yet:

• On Agents' Last Exam, a benchmark of professional computer-use tasks, Terra at medium effort scores 15 percentage points higher than GPT-5.6 Terra at about a third of the cost, and performs on par with GPT-6 Astra at a tenth of its price per token.
• At max effort, it is over three times as token-efficient as Claude Opus 5.5 at a quarter of the price per token, which OpenAI says works out to about 9 times lower cost per task.
• It completes tasks over three times faster than GPT-6 Astra and GPT-6.1 Sol.
• On Terminal-Bench 4.0 at max effort, it solves 50.3% of tasks, versus 21.5% published for GPT-5.6 Terra.
• On an internal benchmark of error-inducing prompts, the error rate at low effort drops from 53.6% with GPT-5.6 Terra to 16.7%; at max effort it is 4.3%.
• It meets the "Critical" cybersecurity threshold under OpenAI's Preparedness Framework and keeps the safeguards built for GPT-6 Astra.

These are the vendor's own numbers, run on the vendor's own setup. Useful for orientation; a run on your own tasks counts for more.

Rumor: that Terra costs $2 input and $10 output. That pair is Terra's Fast mode, or GPT-6.1 Sol at standard.

When it is worth it and when it is not

Worth testing for:

• Repetitive screen work: entering invoice data into a system, filling out forms, reviewing screenshots.
• Reading images with judgment: technical drawings, product photos against a catalog, scanned documents.
• High-volume jobs where cost per task matters more than the last bit of quality, especially if they can run through Batch or Flex.
• Drafting and editing everyday work writing.

Not the first pick for:

• The hardest coding and research work. OpenAI says so itself: GPT-6.1 Sol and GPT-6 Astra remain the better choice there.
• Pipelines that need audio or video in, or images out directly from the model.
• Jobs where latency does not matter but Fast mode is on by default. You pay double for nothing.
• Huge prompts just because "it all fits". Past 272K input tokens the whole request costs more.

Checklist before you switch

1. Budget from the pricing page and the model page, not from a screenshot of them.
2. Check which mode your account or provider runs: standard, Batch, Flex or Fast.
3. Measure how many input tokens your requests use and how many go past 272K.
4. Turn on caching for the prefix that repeats: cached input is 5% of the regular rate.
5. Run 20 or 30 real tasks on your current model and on gpt-6.1-terra, same reasoning effort, and compare output, time and cost.
6. If you plan to use computer use, start in an isolated environment with minimal permissions. The official computer use guide covers how the tool works.
7. For complex code, benchmark Terra against what you already use before switching: OpenAI itself points to other models for that work.

Bottom line

GPT-6.1 Terra's standard price is $1 input, $0.05 cached input and $5 output per million tokens. The $2/$10 going around is Fast mode. What is new and checkable is the combination of a low price, a 1.05M-token window and computer use in the API. The benchmarks come from OpenAI, so test them against your own work.

Frequently asked questions

How much does GPT-6.1 Terra cost in the API?

At standard, $1 per million input tokens, $0.05 per million cached input tokens and $5 per million output tokens, per OpenAI's model page and pricing page.

So does GPT-6.1 Terra cost $2/$10?

Only in Fast mode, which is twice the standard price. It is also the standard price of a different model, GPT-6.1 Sol.

What is GPT-6.1 Terra's context window?

1,050,000 tokens, with up to 922,000 input and 128,000 output tokens. Requests above 272K input tokens are billed at a higher rate.

Is GPT-6.1 Terra better than GPT-6.1 Sol for coding?

According to OpenAI, GPT-6.1 Sol and GPT-6 Astra remain the better choice for the most complex coding. Against GPT-5.6 Terra it does improve, 50.3% versus 21.5% on Terminal-Bench 4.0, also according to OpenAI.

Where can I use GPT-6.1 Terra?

In ChatGPT Work and Codex for Plus, Pro, Enterprise and Edu users, and in the OpenAI API as gpt-6.1-terra.

Sources

• OpenAI: Introducing GPT-6.1 Terra
• OpenAI API: gpt-6.1-terra model page
• OpenAI API: pricing
• OpenAI API: computer use tool guide
• OpenAI on X: availability announcement