GPT-6.1 Terra costs half of Sol: what to move to it

OpenAI released GPT-6.1 Terra on October 26, 2026, at $1 per million input tokens and $5 per million output tokens in the API, half the standard price of GPT-6.1 Sol. The same day it became generally available in GitHub Copilot and on Amazon Bedrock. If you run high-volume, well-scoped work on Sol today, such as document extraction, form filling or routine coding tasks, Terra is the model to test next.

The catch is the word "well-scoped." OpenAI's own announcement says that "for the most complex coding tasks, GPT-6.1 Sol and GPT-6 Astra remain the better choice," and every quality figure in the launch comes from OpenAI's evaluations, not independent tests. A cheaper token only lowers your bill if Terra finishes the task without extra retries.

How much does GPT-6.1 Terra cost?

GPT-6.1 Terra costs $1 per million input tokens, $0.05 per million cached input tokens and $5 per million output tokens at standard API pricing, according to OpenAI's launch post. OpenAI says those prices are "50% lower than those of GPT-5.6 Terra and GPT-6.1 Sol." The API model name is gpt-6.1-terra.

Here is where Terra sits against the other current OpenAI tiers, using the list prices OpenAI published at each launch:

Scroll sideways to see all 4 columns.

Model	Input (per 1M)	Cached input (per 1M)	Output (per 1M)
GPT-6 Luna	$0.10	$0.01	$0.50
GPT-6.1 Terra	$1.00	$0.05	$5.00
GPT-6.1 Sol	$2.00	$0.10	$10.00
GPT-6 Astra	$10.00	$1.00	$50.00

Luna, Sol and Astra prices are from OpenAI's GPT-6.1 Sol announcement of September 29, 2026.

What the price difference looks like on a real workload

Token prices are easier to judge on a monthly job. Take a hypothetical invoice-reading workflow, the kind of high-volume task OpenAI names in its announcement: 100,000 documents a month, each sending about 3,000 input tokens and returning about 400 output tokens. That is 300 million input tokens and 40 million output tokens.

• GPT-6.1 Terra: $300 input + $200 output = $500 a month
• GPT-6.1 Sol: $600 input + $400 output = $1,000 a month
• GPT-6 Luna: $30 input + $20 output = $50 a month
• GPT-6 Astra: $3,000 input + $2,000 output = $5,000 a month

This is arithmetic on list prices, not a measured result. It ignores cached input, which can cut the input line further if your requests reuse a long, stable prefix (see how cached context changed the math for GPT-6.1 Sol), and it ignores any reasoning tokens your requests generate. The figure that decides the question is cost per completed task, retries included.

What OpenAI says Terra does well

OpenAI positions Terra for computer use, visual reasoning and everyday writing. Its claims, all from OpenAI's own evaluations:

• On Agents' Last Exam, a benchmark of professional computer-use tasks, Terra at medium reasoning effort "scores 15 percentage points higher than GPT-5.6 Terra at about one-third the cost," and performs on par with GPT-6 Astra "at only one-tenth its price per token."
• On Terminal-Bench 4.0, Terra at maximum reasoning effort solves 50.3% of tasks, compared with GPT-5.6 Terra's published 21.5%.
• On an internal set of error-inducing prompts, the error rate at low reasoning effort fell from 53.6% with GPT-5.6 Terra to 16.7%.
• OpenAI says Terra completes tasks "over three times faster than both GPT-6 Astra and GPT-6.1 Sol."

GitHub describes the model in similar terms: built "for everyday agentic coding and terminal work," and working "in focused, consolidated steps on well-scoped tasks."

BaristaLabs interpretation: treat these as reasons to run your own comparison, not as a forecast for your workload. The benchmarks measure OpenAI's chosen tasks at OpenAI's chosen reasoning settings.

Where is GPT-6.1 Terra available?

Terra launched in four places on the same day:

• OpenAI API as gpt-6.1-terra.
• ChatGPT Work and Codex for Plus, Pro, Enterprise and Edu users, per OpenAI. The announcement does not name Business plans, so Business admins should check their model list rather than assume access.
• GitHub Copilot for Pro, Pro+, Max, Business and Enterprise users, in the model picker across VS Code, Visual Studio, Copilot CLI, the cloud agent, JetBrains IDEs, Xcode, Eclipse and github.com. GitHub's changelog says the rollout is gradual and the model "is billed at provider list pricing under usage-based billing."
• Amazon Bedrock, generally available with a context window of up to 1 million input tokens and text and image input, per the AWS announcement. Check the Bedrock documentation for supported Regions and Bedrock pricing before you plan on it.

Copilot admins: Terra is on by default

For Copilot Business and Enterprise, GitHub says that under default model enablement "new models are automatically enabled unless an administrator has turned off the global default or explicitly disables this model." If your organization reviews models before developers can pick them, Terra is already in the picker as the rollout reaches you, unless the global default is off.

If you manage that policy in code, this is a good moment to confirm that the change-control process from our post on the Copilot model-policy API actually caught the new model.

Which work should move to Terra, and which should stay on Sol?

BaristaLabs recommendation, based on OpenAI's and GitHub's own positioning:

• Test on Terra: high-volume extraction and form filling, document and image reading, routine agentic coding with a clear acceptance test, and writing drafts. These are the jobs where a 50% price cut adds up and where a wrong answer is cheap to catch.
• Keep on Sol or Astra for now: the hardest multi-file coding work, which OpenAI itself says Sol and Astra still handle better, and any task where a failed run costs more than the tokens.
• Look at Luna first for simple classification or routing, where Terra's extra capability may not earn its 10x higher price.

A first step: rerun one workload on Terra

Pick the single highest-volume workflow you run on GPT-6.1 Sol. Replay a fixed set of 50 to 100 recent, real inputs through gpt-6.1-terra at the same reasoning setting, and record three numbers per model: pass rate against your own acceptance check, total tokens, and retries. If Terra's cost per passed task is lower and its pass rate is within your tolerance, switch that one workflow and watch it for a week. A model bake-off harness makes this repeatable for the next release.

If you want help choosing the workload and setting the acceptance check, our AI consulting service can run the comparison with you.

Sources and scope

• OpenAI, Introducing GPT-6.1 Terra, October 26, 2026. Pricing, availability and all benchmark figures.
• OpenAI, Introducing GPT-6.1 Sol, September 29, 2026. GPT-6 Luna, GPT-6.1 Sol and GPT-6 Astra list prices.
• GitHub Changelog, GPT-6.1 Terra in GitHub Copilot, October 26, 2026. Copilot plans, surfaces, billing and default enablement.
• AWS, OpenAI GPT-6.1 Terra is now generally available on Amazon Bedrock, October 26, 2026. Bedrock availability, context window and input types.

BaristaLabs did not run GPT-6.1 Terra for this article. The cost example is arithmetic on published list prices, and the quality claims are OpenAI's and GitHub's. The sources do not give Bedrock's per-token price or Copilot's per-request cost for Terra.