# GPT-6.1 Terra cuts output price 58%, to $5 per million tokens

> OpenAI's new mid-tier model costs $1 in and $5 out per million tokens. Apps on GPT-5.6 Terra can cut their bill by half or more, after a few checks.

- Published: 2026-10-26
- Author: 28creation Desk
- Format: News
- URL: https://28creation.com/articles/gpt-6-1-terra-price-cut/

## The short read

**OpenAI's GPT-6.1 Terra charges $5 per million output tokens, down from $12 on GPT-5.6 Terra.**

- OpenAI released GPT-6.1 Terra on October 26 at $1 in and $5 out per million tokens, aimed at computer use and writing.

### What you can do with this

> If you run GPT-5.6 Terra, price one week of your logs at the new rates and test your own prompts first. Output-heavy jobs save the most.

---

OpenAI released GPT-6.1 Terra today, its [API changelog](https://developers.openai.com/api/docs/changelog) says, a model built for "precise computer use" and writing. The [model page](https://developers.openai.com/api/docs/models/gpt-6.1-terra) prices it at $1 per million input tokens and $5 per million output. If your app runs on GPT-5.6 Terra, the same traffic now costs half or less. Output fell 58%, more than the promised 50%.

[9to5Mac reported the launch](https://9to5mac.com/2026/10/26/openai-upgrades-chatgpt-and-codex-with-new-gpt-6-1-terra-model/) at 11:21 a.m. Pacific, and the launch announcement appeared in the Announcements section of OpenAI's [developer forum](https://community.openai.com/t/gpt-6-1-terra-delivering-fast-and-capable-computer-use-at-a-lower-price/1409900).

## The output and cache prices fell most

The forum announcement promises "50% lower standard API prices" than GPT-5.6 Terra. That holds for input. The rest of the cut is deeper, going by the [pricing page](https://developers.openai.com/api/docs/pricing):

| Per million tokens | GPT-5.6 Terra | GPT-6.1 Terra | Change |
|---|---|---|---|
| Input | $2.00 | $1.00 | -50% |
| Cached input | $0.20 | $0.05 | -75% |
| Output | $12.00 | $5.00 | -58% |

Those are standard rates for prompts up to 272,000 input tokens. Batch and Flex cost half, $0.50 in and $2.50 out. Terra now also costs exactly half of GPT-6.1 Sol, which lists $2 in and $10 out.

A hypothetical agent that reads 10 million tokens and writes 2 million a month shows the size of it. On GPT-5.6 Terra that is $20 plus $24, or $44. On GPT-6.1 Terra it is $10 plus $10, or $20, a cut of about 55%. The more output and cached input a workload uses, the closer the saving gets to 58% or more.

## Why it matters: Terra now aims at computer use

Terra sits second in OpenAI's four classes, which 9to5Mac ranks by price and capability: Luna at the bottom, Terra next, then Sol, with Astra on top. It was left out of the GPT-6 release last month. The announcement says the model "understands what's on screen" and acts across apps, and the model page lists a computer use tool among the ones it supports, MCP included.

OpenAI's own claims are strong. The announcement says Terra matches GPT-6 Astra "at one-tenth the price per token" and finishes tasks over 3x faster on average than Astra and GPT-6.1 Sol. On OpenAI's internal test of error-inducing prompts, 9to5Mac reports, the error rate at low reasoning effort fell from 53.6% to 16.7%. These are the vendor's numbers; we have not seen an outside benchmark yet.

## The trade-offs to check

The model takes text and images, not audio or video. Tool calling, computer use included, needs the Responses API. The model page lists a single snapshot, `gpt-6.1-terra`, so there is no dated version to pin.

Long prompts cost more. Over 272,000 input tokens, the whole request is billed at 2x input and 1.5x output. Cache writes are charged too: $1.25 per million tokens for 30-minute retention and $2.50 for 12 hours. Knowledge runs to June 1, 2026, against February 16 for GPT-5.6 Terra, whose page lists no shutdown date.

The announcement still points in-depth research and complex coding to Sol or Astra.

## What to do now

1. Price one week of your GPT-5.6 Terra logs at the new rates, with output tokens counted apart from input; output-heavy jobs gain the most.
2. Run your own test set on `gpt-6.1-terra` before switching, since the only benchmarks so far are OpenAI's.
3. Move overnight jobs to Batch or Flex, at $0.50 in and $2.50 out.
4. Flag prompts above 272,000 tokens, where the 2x input rate eats much of the cut.
5. On a paid ChatGPT plan (the announcement names Plus, Pro, Enterprise and Edu), look for Terra in ChatGPT Work and Codex, where the announcement says it is rolling out now and limits stretch further than on Sol or Astra.

## Sources

- OpenAI, API changelog, October 26 entry: https://developers.openai.com/api/docs/changelog
- OpenAI, GPT-6.1 Terra model page: https://developers.openai.com/api/docs/models/gpt-6.1-terra
- OpenAI, GPT-5.6 Terra model page: https://developers.openai.com/api/docs/models/gpt-5.6-terra
- OpenAI, API pricing, read October 26, 2026: https://developers.openai.com/api/docs/pricing
- OpenAI Developer Community, launch announcement: https://community.openai.com/t/gpt-6-1-terra-delivering-fast-and-capable-computer-use-at-a-lower-price/1409900
- 9to5Mac, "OpenAI upgrades ChatGPT and Codex with new GPT-6.1 Terra model": https://9to5mac.com/2026/10/26/openai-upgrades-chatgpt-and-codex-with-new-gpt-6-1-terra-model/