🔍 Read the full analysis: OpenAI’s GPT-6.1 Terra May Get The Job Done For Less on ThorstenMeyerAI.com
Get the little things that make your day delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
TL;DR
OpenAI released GPT-6.1 Terra on Oct. 26, 2026, pricing API use at $1 per million input tokens and $5 per million output tokens. Artificial Analysis found its low- and medium-effort settings on the cost-efficiency frontier in its evaluation, while Terra’s maximum-effort score remained below several larger models. The results are benchmark measurements, not a guarantee of performance or savings for every workload.
OpenAI released GPT-6.1 Terra on Oct. 26, presenting it as a lower-cost model for routine work and computer-based tasks. Its API price is $1 per million input tokens and $5 per million output tokens, while independent benchmark data from Artificial Analysis suggests Terra’s low- and medium-effort settings can offer a better cost-and-score trade-off than tested alternatives. At maximum effort, however, Terra scores below several more capable models.
OpenAI says Terra is available to ChatGPT Work, Codex and API customers, with the API model identifier gpt-6.1-terra. The listed API rates are half the previous Terra model’s $2 input and $12 output rates per million tokens; cached input costs $0.05 per million tokens. OpenAI says eligible ChatGPT plans include Plus, Pro, Enterprise and Edu, and that the same allowance can cover substantially more tasks than use of Sol or Astra. These are the company’s descriptions of availability and value.
Artificial Analysis measured Terra’s maximum-effort score at 48.1 on its Intelligence Index, with an average cost of $0.40 per evaluation task. Its listed comparison figures are $0.72 per task for GPT-6.1 Sol, which scored 51.8, and $5.98 for Claude Opus 5.5, which scored 57.6. Terra therefore cost less in that benchmark, but its score was also lower. The index combines ten evaluations, and the per-task figure is an average across those tasks, not a price applicable to every individual job.
Artificial Analysis tested six effort settings, from non-reasoning to maximum. Its summary places low and medium effort on the cost-efficiency frontier—meaning no tested model scored higher at a lower cost in that comparison. The source material gives the low setting as 36.2 points and $0.055 per task. The full set of figures for all six settings is not available in the supplied material, so the precise results at the other settings cannot be reported here. At the high end, the maximum setting costs about 7.4 times as much as low for 12 additional points, according to the source’s summary.
Infographic 01 · The effort curve
One model, six settings, a 7-fold cost range.
GPT-6.1 Terra at each effort level, as measured by Artificial Analysis. Low and medium sit on its cost-efficiency frontier: no tested model is both cheaper and better.
Score · Intelligence Index
independent higher is better
Cost per finished task
note the last step shorter is cheaper
All six settings
| Effort | Index | Cost / task | Output tokens / task | Hallucination rate |
|---|---|---|---|---|
| Non-reasoning | Index 33.5 | Cost $0.107 | Tokens 4,737 | Hallucination 72% |
| Low | Index 36.2 | Cost $0.055 | Tokens 2,681 | Hallucination 46% |
| Medium | Index 41.3 | Cost $0.095 | Tokens 5,835 | Hallucination 44% |
| High | Index 43.9 | Cost $0.134 | Tokens 9,027 | Hallucination 43% |
| Xhigh | Index 45.4 | Cost $0.194 | Tokens 14,551 | Hallucination 42% |
| Max | Index 48.1 | Cost $0.403 | Tokens 35,096 | Hallucination 38% |
Hallucination rate in AA-Omniscience: how often the model answers anyway when it does not know. Lower is better.
The first steps up are cheap. The last one doubles the bill.Assessment: Thorsten Meyer
Sources: Artificial Analysis model pages (retrieved 26 Oct 2026); Artificial Analysis launch article (26 Oct 2026). Cost per task is the average across the index’s evaluation tasks.
Powered by Thorsten Meyer AI
Why Cost per Task Matters
Terra’s launch puts a practical question ahead of headline model rankings: how much does a usable result cost? Token prices alone do not answer that. The source reports that GPT-6 Luna is ten times cheaper than Terra per token, yet Terra at low effort costs less per completed task in Artificial Analysis’s measurement because it produces fewer tokens. A model that charges less per token can still be more expensive to use if it needs more output or reasoning to finish a job.
For businesses processing high volumes of routine work, such as invoice reading and data entry, the benchmark offers a reason to test a smaller model before paying for a larger one. But the results do not show that Terra will be the least expensive option for every company: actual costs depend on workload, prompt design, output length, settings and the level of accuracy required. OpenAI says Terra completes tasks more than three times faster than Astra and Sol, but Artificial Analysis had not published speed measurements when the source checked its data.
There is also a trade-off in capability. Artificial Analysis scored Terra below Claude Opus 5.5, Kimi K3.1 Pro, GPT-6 Astra and GPT-6.1 Sol at the listed settings. The source also notes that open-weight MiMo-V2.6-Pro scored 46.3 for $0.13 per task, below Terra’s maximum score but at a lower measured cost. The benchmark therefore supports testing models by task and total cost, rather than treating Terra as a universal replacement.
Products Worth Considering
Withings Sleep Under-Mattress Sleep Tracking Pad
Garmin Forerunner 570 GPS Running Smartwatch, 47mm, Aluminum, Black
Oura Ring 5 Smart Ring – Silver – Size 10
As an affiliate, we earn on qualifying purchases.
Terra’s Place in OpenAI’s Lineup
OpenAI describes Terra as a fast, efficient model for communication and computer-based tasks that fill a workday, including processing incoming invoices. It sits between the smaller GPT-6 Luna and the larger Sol and Astra models. The supplied launch material says its API prices are also half those of GPT-6.1 Sol, at $2 input and $10 output per million tokens, and GPT-5.6 Terra, at $2 and $12 respectively.
OpenAI’s own published comparisons and Artificial Analysis’s independent evaluation are separate evidence and should not be combined as if they were one test. OpenAI reports that Terra at medium effort scored 57.6% on Agents’ Last Exam for $0.29 per task, matching GPT-6 Astra’s score there at a reported $4.82 per task. At maximum effort, OpenAI says Terra cost $1.01 per task versus $8.82 for Claude Opus 5.5 on that test; its chart also shows Opus scoring 63.2% and Terra 60.0%. Those figures are OpenAI’s benchmark claims, while Artificial Analysis used its own index and cost methodology.
OpenAI also reports improved results against its previous Terra model on several internal or published measures: a lower factual-error rate on its internal error-inducing prompts, and a 50.3% result on Terminal-Bench 4.0 at maximum effort compared with a published 21.5% for GPT-5.6 Terra. OpenAI says Sol and Astra remain better choices for the most complex coding tasks. It also classifies Terra as meeting the “Critical” threshold in cybersecurity under its Preparedness Framework; that classification is OpenAI’s assessment, not an independent finding in the material provided.
What the Benchmarks Cannot Show
The reported scores are benchmark results, not guarantees about real-world outcomes. The supplied material does not provide the full task-by-task breakdown, confidence intervals, or enough detail to establish how closely the benchmark workloads match a particular customer’s use. Cost per task is an average across Artificial Analysis’s evaluation set; a company’s own cost could differ.
Some comparisons are also based on different tests. OpenAI’s Agents’ Last Exam figures and Artificial Analysis’s Intelligence Index results measure different evaluations, so their scores and per-task costs should not be read as directly interchangeable. Artificial Analysis had not published speed measurements for Terra when the source checked on launch day, leaving OpenAI’s claim that Terra is more than three times faster than Astra and Sol independently unverified in the material available.
The supplied source says Terra guessed on 38% of questions it could not answer in Artificial Analysis’s hallucination test, compared with 88% for GPT-5.6 Terra. That is a result within a particular test; it does not establish the rate of incorrect answers across all uses. The available source material also does not give all six effort-setting scores and costs, preventing a complete review of the curve.
Testing Terra on Real Workloads
The immediate next step for customers is to compare Terra with larger and lower-cost alternatives on their own tasks. A useful evaluation would track cost per accepted result, completion time and error rates at each effort setting, rather than relying only on token prices or a single benchmark score. For routine work, low or medium effort is the starting point suggested by Artificial Analysis’s cost-efficiency results; more demanding tasks may justify a larger model.
Further independent speed measurements and fuller task-level results would help test OpenAI’s performance claims and clarify how Terra’s six effort settings behave across different kinds of work. Until those data are available, the launch establishes Terra’s pricing and benchmark position, but not the savings a particular user or organization will achieve.
Key Questions
How much does GPT-6.1 Terra cost?
OpenAI lists API prices of $1 per million input tokens and $5 per million output tokens. Cached input is listed at $0.05 per million tokens. Artificial Analysis measured maximum-effort use at an average $0.40 per task across its evaluation set; that is a benchmark average, not a universal task price.
Is Terra more capable than OpenAI’s larger models?
Not according to the supplied Artificial Analysis comparison at maximum effort. Terra scored 48.1, below GPT-6.1 Sol at 51.8 and GPT-6 Astra at 52.7. OpenAI also says Sol and Astra remain better choices for its most complex coding tasks.
Why compare cost per task instead of token prices?
Token rates do not show how many tokens a model uses to complete a job. In Artificial Analysis’s measurements, Terra at low effort cost less per task than GPT-6 Luna despite Luna’s lower token price, because Terra produced fewer tokens in the tested tasks.
What did OpenAI say about Terra’s performance?
OpenAI reported improvements on its computer-use, coding and internal factual-error measures compared with GPT-5.6 Terra. Those are company-reported results, separate from Artificial Analysis’s independent index, and may not predict performance on every workload.
What remains unknown about Terra?
The supplied material does not establish customer savings across real-world workloads or provide independent speed measurements. It also lacks complete results for all six effort settings and full task-by-task benchmark details.
Source: ThorstenMeyerAI.com
