🔍 Read the full analysis: GPT-6.1 Terra: How Much Intelligence Does The Job Actually Need? on ThorstenMeyerAI.com
TL;DR
OpenAI released GPT-6.1 Terra on October 26, 2026, positioning it as a lower-priced model for routine workplace and computer-based tasks. Artificial Analysis measured its cost and benchmark performance across six effort settings: low and medium offered the strongest cost-to-score trade-off in its tests, while higher settings cost more and still scored below several larger models.
OpenAI released GPT-6.1 Terra on October 26, describing it as a fast, efficient model for workplace communication and computer-based tasks. Its API price is $1 per million input tokens and $5 per million output tokens, half the listed rates for GPT-5.6 Terra, while independent testing found that Terra’s value depends on the effort setting and the cost of completing a task—not just token prices.
Terra is available through the API as gpt-6.1-terra, and to Plus, Pro, Enterprise and Edu users in ChatGPT Work and Codex, according to OpenAI. The company says its usage allowance can cover substantially more tasks with Terra than with Sol or Astra. OpenAI presents invoice processing—reading incoming invoices and entering their details into accounting software—as an example of the high-volume work it targets.
Artificial Analysis, an independent benchmarking firm, gave Terra an Intelligence Index score of 48.1 at maximum effort, with an average evaluation-task cost of $0.40. Its figures put GPT-6.1 Sol at 51.8 and $0.72 per task, Claude Opus 5.5 at 57.6 and $5.98, and GPT-6 Luna at 38.1 and $0.07. These scores and costs reflect the firm’s evaluation set; they are not a guarantee of performance or spending on a particular customer’s workload.
Across Terra’s six effort settings, Artificial Analysis recorded scores ranging from 33.5 to 48.1 and costs from $0.055 to $0.40 per task. Low and medium effort were on the firm’s cost-efficiency frontier, meaning no model it tested was both cheaper and better at those points. At maximum effort, Terra scored 12 points above its low setting but cost about 7.4 times as much. The non-reasoning setting was not the least expensive: it cost $0.107 per task, compared with $0.055 at low effort, and scored lower.
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
Task Costs Change the Model Choice
The launch puts attention on a practical distinction for teams buying AI services: token price is not the same as cost per completed task. A cheaper model can use more tokens or need more attempts, while a model with higher per-token rates may finish work with fewer resources. Artificial Analysis reported that Luna, despite its lower token prices, cost more per task than Terra at low effort in its evaluation. That result applies to the tasks and measurement method used by the firm, not every real-world workflow.
For organizations handling large volumes of routine work, Terra’s lower listed rates and low-to-medium effort results could make it worth testing against more capable models. The trade-off is that Terra does not lead the cited intelligence rankings: Artificial Analysis placed Opus 5.5, Kimi K3.1 Pro and Gemini 4 Argon above it at the listed settings, and said Xiaomi’s open-weight MiMo-V2.6-Pro scored higher than Terra’s high and extra-high settings while costing less per evaluation task. OpenAI also says Sol and Astra remain better choices for the most complex coding work.
The figures point to routing by task difficulty as a possible deployment approach: use a less costly setting for routine jobs and reserve stronger models for work where errors or weak reasoning carry greater costs. That is an operational interpretation, not a recommendation validated across businesses. Companies would need to test their own tasks, including review time, failure rates and any costs not captured by benchmark averages.
Popular right now
Ninja Foodi Air Fryer 10 Qt DualZone 6-in-1 Black DZ302
Ninja DualZone Smart Air Fryer, 10 Qt, 6-in-1, DZ950
KitchenAid Classic Series 4.5-Quart Tilt-Head Stand Mixer, Onyx Black
As an affiliate, we earn on qualifying purchases.
Where Terra Fits in OpenAI’s Lineup
OpenAI places Terra between GPT-6 Luna, its smaller model, and the larger Sol and Astra models. Its listed API rates are also below GPT-6.1 Sol’s $2 per million input tokens and $10 per million output tokens, and GPT-6 Astra’s $10 and $50 rates, as shown in the supplied launch material. Cached input for Terra is listed at $0.05 per million tokens.
OpenAI’s launch page reports results from its own evaluations as well as speed and safety claims. On Agents‘ Last Exam, the company says Terra at medium effort scored 57.6% and cost $0.29 per task, compared with GPT-5.6 Terra at 42.6% and $0.84. OpenAI also reports that Terra solved 50.3% of Terminal-Bench 4.0 tasks at maximum effort, compared with 21.5% for GPT-5.6 Terra’s published result. The figures come from OpenAI’s own tests and should not be conflated with Artificial Analysis’s independent index.
OpenAI says Terra completes tasks more than three times faster than Astra and Sol, but Artificial Analysis had not published speed measurements for Terra at the time of the source report. OpenAI also classifies Terra as meeting the Critical threshold in cybersecurity under its Preparedness Framework and says it retains safeguards developed for GPT-6 Astra. Those are company statements; the supplied material does not provide an independent verification of them.
Limits of the Benchmark Evidence
The available material does not establish how Terra will perform across customers‘ real workloads, or whether the benchmark cost averages will match their usage. Task mix, prompt length, retries and human review can affect practical costs, and the cited figures do not quantify those factors for individual deployments.
The source provides only part of the six-setting results table: it gives non-reasoning, low and maximum scores and costs, but not every setting’s full metrics. It says low and medium sit on the cost-efficiency frontier, but the complete underlying comparison is not included here. Artificial Analysis had also not published Terra speed measurements when the source was written, leaving OpenAI’s speed comparison unverified by that firm.
OpenAI’s own benchmark, factual-error and cybersecurity statements are company claims. The supplied information does not include independent replication, detailed test methods for every claim, or evidence about how the safety classification translates to particular deployments. It is also unclear how quickly prices, model availability or benchmark standings may change after launch.
Testing Terra on Real Workloads
The next useful measure will be how Terra performs in customer deployments and in further independent evaluations. Teams considering it can compare cost per successfully completed task across Terra’s effort settings and relevant alternatives, while tracking errors, completion time and human review needs. Benchmark scores alone cannot determine which setting is suitable for a specific workflow.
OpenAI has made Terra available through its API, ChatGPT Work and Codex, so users can assess it within those products. The source material does not give a schedule for additional model releases or updated independent tests. Readers should treat the October 26 figures as a launch-day snapshot rather than a lasting ranking.
Key Questions
What is GPT-6.1 Terra?
GPT-6.1 Terra is an OpenAI model positioned between Luna and the larger Sol and Astra models. OpenAI says it is aimed at workplace communication and computer-based tasks.
How much does GPT-6.1 Terra cost through the API?
The listed price is $1 per million input tokens and $5 per million output tokens, with cached input at $0.05 per million tokens, according to the launch material. Actual costs depend on usage and task requirements.
Which Terra effort setting is most cost-efficient?
In Artificial Analysis’s evaluation, low and medium effort were on the cost-efficiency frontier. The supplied source does not include the full figures for every setting, and the best choice for a particular job may differ.
Does Terra outperform OpenAI’s larger models?
Not across the cited measures. Artificial Analysis scored Terra below GPT-6.1 Sol, and OpenAI says Sol and Astra remain better for the most complex coding tasks. Terra’s case is its lower price and performance on selected tasks, not a claim that it leads every benchmark.
Where can users access GPT-6.1 Terra?
OpenAI says Terra is available through the API as gpt-6.1-terra and to Plus, Pro, Enterprise and Edu users in ChatGPT Work and Codex.
Source: ThorstenMeyerAI.com
