OpenAI released GPT-6.1 Terra on October 26, 2026, a model it calls “a fast, efficient model that combines detailed visual reasoning and precise computer use alongside clearer, more natural writing”. It is in the API as gpt-6.1-terra and in ChatGPT Work and Codex for Plus, Pro, Enterprise and Edu users from day one.
The headline is price. At $1 per million input tokens and $5 per million output, Terra costs half of GPT-6.1 Sol, and on OpenAI’s own agent benchmark it scores as high as GPT-6 Astra at medium effort for about one-seventeenth of the cost per task.
On this page · 8 sectionsOpen
OpenAI released GPT-6.1 Terra on October 26, 2026, in the API as gpt-6.1-terra and in ChatGPT Work and Codex for Plus, Pro, Enterprise and Edu users.
Standard API pricing is $1 per million input tokens, $0.05 cached and $5 output, half of GPT-6.1 Sol’s $2 and $10 and down from GPT-5.6 Terra’s $2 and $12.
On OpenAI’s Agents’ Last Exam chart, Terra at medium effort scores 57.6%, the same as GPT-6 Astra at medium, for $0.29 a task against Astra’s $4.82.
At max effort Terra scores 60.0% for $1.01 a task; Claude Opus 5.5 scores 63.2% for $8.82, so Opus still leads that chart at about 9x the cost.
On Terminal-Bench 4.0 at max effort Terra solves 50.3% of tasks against GPT-5.6 Terra’s 21.5%, and OpenAI says it meets the Critical cybersecurity threshold in its Preparedness Framework.
§ 01What OpenAI released
Terra takes GPT-5.6 Terra’s place in OpenAI’s four-model line: GPT-6 Astra for the hardest work, GPT-6.1 Sol for balanced work, Terra for fast and efficient work, and GPT-6 Luna for high-volume focused tasks.
- Inputs and outputs. Text and images in, text out. Audio and video are unsupported.
- Context. 1,050,000 tokens, up to 922,000 of them input, and 128,000 output tokens.
- Knowledge cutoff. June 1, 2026, the most recent in OpenAI’s current lineup.
- Reasoning effort. none, low, medium (the default), high, xhigh and max. Astra and Sol do not offer none.
- Tools. Chat Completions works, but tool calling requires the Responses API, where Terra supports computer use, web search, file search, code interpreter, hosted shell, MCP and skills.
- Where it runs. ChatGPT Work and Codex for Plus, Pro, Enterprise and Edu; the announcement does not name Free or Business. OpenAI says the same plan allowance covers more tasks with Terra than with GPT-5.6 Terra, and substantially more than with Sol or Astra.
§ 02Price: half of Sol, under half of GPT-5.6 Terra
| Model | Input | Cached input | Output | Context | Knowledge cutoff |
|---|---|---|---|---|---|
| GPT-6.1 Terra | $1 | $0.05 | $5 | 1.05M | Jun 1, 2026 |
| GPT-5.6 Terra | $2 | $0.20 | $12 | 1.05M | Feb 16, 2026 |
| GPT-6.1 Sol | $2 | $0.10 | $10 | 1.05M | Apr 30, 2026 |
| GPT-6 Astra | $10 | $1.00 | $50 | 1.05M | Apr 30, 2026 |
| GPT-6 Luna | $0.10 | $0.01 | $0.50 | 1.05M | May 18, 2026 |
The fine print on Terra’s model page:
- Cache writes cost $1.25 per million tokens with 30-minute retention and $2.50 with 12-hour retention.
- Prompts over 272,000 input tokens bill at 2x input and cache rates and 1.5x output for the whole request.
- Batch and Flex are 50% cheaper than Standard; Fast mode is 2x; regional processing adds 10%.
Output is where the cut is deepest: $5 against GPT-5.6 Terra’s $12 is 58% less, and output is most of the bill for agents that write long answers or code.
§ 03Benchmarks: Agents’ Last Exam
OpenAI measured computer use on Agents’ Last Exam, a set of professional tasks such as turning architectural drawings into a 3D model. Its chart plots score against estimated cost per task for six models at each reasoning effort. We read the data points from the chart on OpenAI’s page.
| Model | Effort | Score | Cost per task |
|---|---|---|---|
| GPT-6.1 Terra | Low | 55.9% | $0.16 |
| GPT-6.1 Terra | Medium | 57.6% | $0.29 |
| GPT-6.1 Terra | Max | 60.0% | $1.01 |
| GPT-5.6 Terra | Medium | 42.6% | $0.84 |
| GPT-5.6 Terra | Max | 50.7% | $3.58 |
| GPT-6.1 Sol | Medium | 53.8% | $1.31 |
| GPT-6.1 Sol | Max | 58.6% | $2.00 |
| GPT-6 Astra | Medium | 57.6% | $4.82 |
| GPT-6 Astra | Max | 59.3% | $7.24 |
| GPT-6 Luna | Max | 50.9% | $0.14 |
| Claude Opus 5.5 | Max | 63.2% | $8.82 |
What the chart says:
- Terra beats Sol at every effort level on this test. 55.9% against 52.4% at low, 57.6% against 53.8% at medium, 60.0% against 58.6% at max, each for less money per task.
- Terra at medium ties Astra at medium. Both score 57.6%; Terra’s estimated cost is $0.29 a task against Astra’s $4.82.
- Opus 5.5 still scores highest. 63.2% at max effort against Terra’s 60.0%. OpenAI’s line is that Terra gets there at “a ~9x lower cost per task”, which matches $8.82 against $1.01 on its own chart.
- The jump from GPT-5.6 Terra is 15 points at medium. 57.6% against 42.6%, at about a third of the cost.
These are OpenAI’s runs, on OpenAI’s chart, with OpenAI’s cost estimates. No outside lab has published Terra results yet.
§ 04The other numbers OpenAI published
- Terminal-Bench 4.0, max effort: 50.3% of tasks solved, against GPT-5.6 Terra’s published 21.5%.
- Errors on OpenAI’s internal set of error-inducing prompts: at low effort, from 53.6% with GPT-5.6 Terra to 16.7%; at max effort, down 18 points to 4.3%.
- Terminal-Bench-Science 0.1, max effort: OpenAI says Terra beats Sol at lower cost and comes within five points of Opus 5.5 at less than one-fifth of the estimated cost. No scores in the text.
- MMMU-Pro and OSWorld 2.1: OpenAI says Terra beats GPT-5.6 Terra at every reasoning setting on MMMU-Pro and passes GPT-5.6 Terra’s best OSWorld 2.1 score at medium effort. No scores in the text.
- Safety: OpenAI says Terra “meets the Critical threshold in cybersecurity under our Preparedness Framework” and keeps the protections built for GPT-6 Astra.
§ 05What OpenAI left out
- Hard coding. OpenAI says it plainly: “For the most complex coding tasks, GPT-6.1 Sol and GPT-6 Astra remain the better choice.”
- Scores for half its claims. MMMU-Pro, OSWorld 2.1 and Terminal-Bench-Science are described in words on the page we read, not numbers.
- Free and Business plans. The announcement lists Plus, Pro, Enterprise and Edu for ChatGPT Work and Codex.
- A launch post in OpenAI’s news feed. At 18:04 UTC on launch day the announcement page and the API changelog were live, but OpenAI’s RSS feed did not list it yet.
§ 06Our conflict, and what it means for a builder
Our conflict, declared: every CellCog tier runs on Claude Opus 5.5, the model that tops this chart and costs the most on it. We do not route to Terra.
For anyone automating high-volume screen and document work, Terra changes the default math. OpenAI’s own example is reading invoices and typing them into accounting software, work where an agent repeats one task thousands of times and the cost per task matters more than the last three points of score. Test it on your own tasks at medium effort before paying for max.
§ 07What we are watching
- Outside benchmarks. Independent runs of Terra on agent and coding boards, against OpenAI’s own chart.
- The system card. OpenAI points to it for the safety detail behind the Critical cybersecurity rating.
- Cloud availability. Whether Terra joins Astra, Sol and Luna on Google Cloud and Microsoft’s platforms.
- Plan coverage. Whether Free and Business users get Terra in ChatGPT.
§ 08Sources
- OpenAI, Introducing GPT-6.1 Terra, October 26, 2026, read 18:04 UTC, chart data points read from the page.
- OpenAI, API changelog, entry for October 26, 2026.
- OpenAI, GPT-6.1 Terra model page, specs and pricing, read October 26, 2026.
- OpenAI, API pricing, read October 26, 2026.
- OpenAI, GPT-6 guide, model lineup and migration notes, read October 26, 2026.
- OpenAI model pages for GPT-5.6 Terra, GPT-6.1 Sol, GPT-6 Astra and GPT-6 Luna.
Q1When was GPT-6.1 Terra released?
October 26, 2026. OpenAI’s API changelog lists it under that date, and its announcement says it is available starting today in ChatGPT Work, Codex and the API.
Q2How much cheaper is GPT-6.1 Terra than GPT-5.6 Terra?
Input drops from $2 to $1 per million tokens, cached input from $0.20 to $0.05, and output from $12 to $5. OpenAI rounds this to standard API prices 50% lower than GPT-5.6 Terra and GPT-6.1 Sol.
Q3Does GPT-6.1 Terra beat Claude Opus 5.5?
Not on score. On OpenAI’s Agents’ Last Exam chart Opus 5.5 at max effort scores 63.2% against Terra’s 60.0%. Terra’s case is cost: $1.01 a task against $8.82. These are OpenAI’s own runs.
Q4What is GPT-6.1 Terra's context window?
1,050,000 tokens, with up to 922,000 input tokens and 128,000 output tokens. Its knowledge cutoff is June 1, 2026, the latest of OpenAI’s current models.
Q5Can GPT-6.1 Terra take audio or video?
No. It takes text and images and returns text. Audio and video are listed as unsupported, and tool calling requires the Responses API.