OpenAI Releases GPT-6.1 Terra, a Computer Use Model Priced at Half of GPT-6.1 Sol

OpenAI on October 26, 2026 released GPT-6.1 Terra, a mid-priced model for computer use, visual reasoning and writing, in ChatGPT Work, Codex and its API. Its standard API price of $1 per million input tokens and $5 per million output tokens is 50% lower than GPT-5.6 Terra and GPT-6.1 Sol, the company said.

Terra replaces GPT-5.6 Terra, which OpenAI released in July 2026 with the rest of its GPT-5.6 family (DataNorth), and sits between the larger Sol and Astra models and the smaller GPT-6 Luna. It follows other new models this month that compete mostly on price, such as Anthropic's Claude Haiku 5.5 on October 7. The Information framed the launch as part of a pricing battle with Anthropic. A day earlier, on October 25, OpenAI was reported to be offering steep discounts to win Anthropic customers such as Replit (The Information via The Decoder).

What did OpenAI release?

GPT-6.1 Terra, released on October 26, 2026, is what OpenAI calls a "fast, efficient model" for office work such as drafting proposals, reading technical drawings, filling out forms and editing videos. Computer use means the model reads what is on a screen, then clicks, types and moves between applications to finish a task.

In the API the model is called gpt-6.1-terra. OpenAI's model page lists a 1,050,000 token context window, up to 128,000 output tokens, a knowledge cutoff of June 1, 2026 and six reasoning effort settings from none to max, with medium as the default.

In ChatGPT, Terra is available to Plus, Pro, Enterprise and Edu users in ChatGPT Work and Codex. OpenAI says a plan's usage allowance covers more tasks with Terra than with GPT-5.6 Terra, Sol or Astra.

How much cheaper is GPT-6.1 Terra?

At $1 and $5 per million input and output tokens, GPT-6.1 Terra costs exactly half of GPT-6.1 Sol's $2 and $10, according to OpenAI's model pages. Against GPT-5.6 Terra, at $2 and $12, input is half the price and output works out to about 58% cheaper.

Cached input, meaning repeated prompt text the API has already processed, costs $0.05 per million tokens, a quarter of GPT-5.6 Terra's $0.20.

The comparison with Anthropic is just as direct. At list prices, Terra costs half as much per token as Claude Sonnet 5.5, the Anthropic model The Deep View compared it with, and a quarter as much as Claude Opus 5.5, according to Anthropic's price list. Price per token is only part of the cost of a task, because models use different numbers of tokens for the same work.

Model	Input	Output	Cached input
GPT-6.1 Terra	$1	$5	$0.05
GPT-5.6 Terra	$2	$12	$0.20
GPT-6.1 Sol	$2	$10	$0.10
GPT-6 Luna	$0.10	$0.50	$0.01
Claude Sonnet 5.5	$2	$10	$0.10
Claude Opus 5.5	$4	$20	$0.20

How does Terra perform, according to OpenAI?

OpenAI says GPT-6.1 Terra at medium reasoning effort scores 15 percentage points higher than GPT-5.6 Terra on Agents' Last Exam, a benchmark of professional computer use tasks, at about one third of the cost. On the same test, the company says Terra performs on par with GPT-6 Astra at one tenth of Astra's price per token and finishes tasks over three times faster than Astra and GPT-6.1 Sol.

At max effort, OpenAI says Terra is over three times as token efficient as Claude Opus 5.5 at a quarter of its price per token, for roughly nine times lower cost per task. The Deep View points out that Opus 5.5 is a previous generation model; Anthropic's price list also includes Claude Fable 5.1, at $10 and $50 per million tokens.

OpenAI's other figures are also self reported:

• On an internal set of prompts built to induce factual errors, the error rate at low effort fell from 53.6% with GPT-5.6 Terra to 16.7%, and at max effort it fell 18 points to 4.3%.
• On Terminal-Bench 4.0, a test of coding agents working from the command line, Terra solves 50.3% of tasks at max effort, against 21.5% for GPT-5.6 Terra.

OpenAI still recommends GPT-6.1 Sol and GPT-6 Astra for the most complex coding tasks.

What do independent measurements show?

Artificial Analysis, an independent benchmarking company, reported on October 26, 2026 that GPT-6.1 Terra at max effort scores 6 points higher than GPT-5.6 Terra on its Intelligence Index, at $0.40 per task against $1.40. On that index Terra sits 4 points below GPT-6.1 Sol and 10 points above GPT-6 Luna, according to its analysis.

On the firm's new Coding Agent Index v2.0, Terra scores 48, against 54 for Sol and 31 for Luna. Its hallucination rate on AA-Omniscience, meaning how often it guesses when it does not know an answer, fell to 38% from 88% for GPT-5.6 Terra, while accuracy rose 2 points.

Artificial Analysis measured a Terminal-Bench 4.0 gain of about 20 points over GPT-5.6 Terra, smaller than the gap of about 29 points in OpenAI's own figures, though the two may use different settings. It called gains on most other evaluations small.

What did OpenAI say about safety?

OpenAI says GPT-6.1 Terra meets the Critical threshold in cybersecurity under its Preparedness Framework, the system it uses to rate dangerous capabilities, and keeps the protections it developed for GPT-6 Astra. The company also says Terra is more likely than GPT-5.6 Terra to stay within the scope of a task and disclose its limitations, and that it resisted prompt injection, hidden instructions meant to divert it from the user's task, in OpenAI's tests.

The scope claim stands out because doubts about staying "within scope and authorization," in the words of OpenAI head of safety systems Saachi Jain, led OpenAI to shelve the launch of GPT-6.1 Astra on September 28, 2026, according to The Hacker News.

What we don't know yet

• Whether Terra can be used with data processing inside the EU; OpenRouter lists it as not available through EU in-region routing as of October 26, 2026.
• How Terra compares with Claude Sonnet 5.5 on independent computer use tests.
• What the Critical cybersecurity rating changes in practice for API customers beyond the Astra safeguards.
• When OpenAI plans to retire GPT-5.6 Terra.