🔍 Read the full analysis: How To Get Everyday AI Tasks Done For Less With GPT-6.1 Terra on ThorstenMeyerAI.com
Get the latest gadgets 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 measured an average task cost of $0.40 at maximum effort and found low and medium settings offer better cost-efficiency, while higher-scoring models remain available for harder work.
OpenAI released GPT-6.1 Terra on Oct. 26, positioning it between its smaller Luna model and larger Sol and Astra models as a lower-cost option for routine work. Its API rates are $1 per million input tokens and $5 per million output tokens, half the listed rates for GPT-5.6 Terra; independent testing by Artificial Analysis suggests the more useful comparison is cost per completed task, which ranges from about $0.05 to $0.40 across Terra’s reasoning settings.
OpenAI describes Terra as a fast, efficient model for communication and computer-based work, citing invoice reading and data entry as examples. The company says the model is available through the API as gpt-6.1-terra and to Plus, Pro, Enterprise and Edu users in ChatGPT Work and Codex. OpenAI says those users can complete substantially more tasks within the same allowance than with Sol or Astra; that allowance comparison is the company’s claim.
Artificial Analysis’s launch-day evaluation gave Terra an Intelligence Index score of 48.1 at maximum effort, at an average cost of $0.40 per task across the index’s evaluations. It scored below GPT-6.1 Sol at 51.8, Claude Opus 5.5 at 57.6 and several other models, but above GPT-6 Luna at 38.1. Artificial Analysis said Terra’s maximum-effort result was six points above GPT-5.6 Terra’s score, at less than a third of that model’s measured task cost.
Across six effort settings, Artificial Analysis reported scores from 33.5 to 48.1 and task costs from roughly $0.05 to $0.40. Its measurements put low and medium effort on the cost-efficiency frontier: tested alternatives did not provide a better score for a lower cost in those comparisons. The source material says maximum effort costs 7.4 times as much as low effort for 12 additional index points. A non-reasoning setting was not the cheapest: it cost nearly twice as much per task as low effort while scoring 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
Why Task Costs Beat Token Rates
Terra’s pricing highlights why a low price per token does not necessarily mean a low bill for completed work. A model may use more tokens to finish a task, while another may produce a shorter response or require less reasoning. Artificial Analysis measured Terra at low effort costing less per task than Luna, even though Luna is ten times cheaper per token, because Terra used substantially fewer output tokens in the cited comparison.
For organizations processing large volumes of routine tasks, the distinction affects budgeting and model selection. The figures support testing a lower effort setting on repeatable work and measuring the cost per successful completion, rather than choosing solely by API rates. They do not show that Terra is the best choice for every task: its score trails several larger models, and OpenAI says Sol and Astra remain better choices for its most complex coding work.
For individual users, lower rates may stretch paid usage allowances, but the actual benefit depends on the work being done, how often the model needs retries, and whether its answers meet the user’s quality requirements. The benchmark averages are useful comparisons, not a guarantee of an individual’s costs or results.
Products Worth Considering

Apple MacBook Air 13-inch Laptop with M5 Chip, 16GB Memory, 512GB SSD, Liquid Retina Display, and 12MP Center Stage Camera (Midnight)

Apple Watch Series 12 GPS 42mm Smartwatch

DJI Mini 3 Drone with DJI RC Controller
As an affiliate, we earn on qualifying purchases.
Terra Between Luna and Sol
OpenAI’s launch places Terra in a tier between Luna and the larger Sol and Astra models. The listed API prices are 50% lower than GPT-5.6 Terra’s: the earlier model was priced at $2 per million input tokens and $12 per million output tokens. Terra’s cached-input price is listed as $0.05 per million tokens.
OpenAI also reported results from its own evaluations. It said Terra at medium effort scored 57.6% on Agents’ Last Exam, compared with 42.6% for GPT-5.6 Terra at medium effort, at a reported cost of $0.29 rather than $0.84 per task. On Terminal-Bench 4.0, OpenAI reported a 50.3% result for Terra at maximum effort, compared with 21.5% for GPT-5.6 Terra. These are company-provided benchmark results, distinct from Artificial Analysis’s independent index.
OpenAI also reported lower error rates on its internal set of error-inducing prompts and said Terra completes tasks more than three times faster than Astra and Sol. Artificial Analysis had not published speed measurements for Terra at the time described in the source. OpenAI’s claim that Terra meets the “Critical” threshold in cybersecurity under its Preparedness Framework is also the company’s own classification, not an independent finding in the cited material.
Limits of the Benchmark Evidence
The published figures do not establish how Terra will perform across every real-world workload. Artificial Analysis’s task costs are averages across its evaluation set, and results can differ with task length, prompt design, tool use, retries and the effort setting selected. The supplied material does not provide the complete values for every row of Terra’s six-setting curve, so comparisons beyond the listed figures should be treated cautiously.
OpenAI’s speed, error-rate, computer-use and cybersecurity statements rely on company data or classifications. Independent speed measurements were not available in the cited material. It is also unclear from the launch information how ChatGPT allowance usage will translate into a fixed number of tasks, because task complexity and consumption vary.
Terra’s score is below several models in Artificial Analysis’s index, and the benchmark also lists cheaper alternatives, including the open-weight MiMo-V2.6-Pro. Those rankings do not, on their own, establish which model is most suitable for a particular company’s tools, privacy requirements or accuracy needs.
Testing Terra on Real Workloads
The next useful test for buyers is to run Terra against representative tasks from their own workflows, beginning with low or medium effort and recording both completion quality and total cost. Teams can compare those results with Sol, Astra or other available models before routing more demanding work to a higher tier. That approach reflects the evidence available so far: Terra’s lower token rates and benchmark costs are documented, but the best setting depends on the task.
Further independent measurements, particularly speed results and evaluations across practical business workflows, could clarify how the launch-day benchmark translates to day-to-day use. OpenAI’s published rates and availability establish the current offer; actual user spending and performance will depend on usage and task mix.
Key Questions
How much does GPT-6.1 Terra cost through the API?
OpenAI’s listed rate is $1 per million input tokens and $5 per million output tokens. Cached input is listed at $0.05 per million tokens.
Which Terra effort setting is the best value?
Artificial Analysis’s evaluation placed low and medium effort on its cost-efficiency frontier. Low effort had the lowest reported task cost among the cited settings; the right choice still depends on the accuracy and reasoning a task requires.
Is Terra better than GPT-6.1 Sol or Claude Opus 5.5?
Not by the cited maximum-effort Intelligence Index scores. Terra scored 48.1, compared with 51.8 for Sol and 57.6 for Claude Opus 5.5, though Artificial Analysis measured lower average task costs for Terra than for those models.
Where can users access GPT-6.1 Terra?
OpenAI says Terra is available through the API under the model name gpt-6.1-terra and to Plus, Pro, Enterprise and Edu users in ChatGPT Work and Codex.
What remains unknown about Terra’s performance?
Independent speed measurements were not available in the cited launch-day material, and benchmark averages cannot predict results for every workflow. Real-world costs will depend on task complexity, effort settings, retries and how well the model performs the work.
Source: ThorstenMeyerAI.com
