🔍 Read the full analysis: GPT-6.1 Terra: A Lower-Cost Model For Work That Doesn’t Need The Best on ThorstenMeyerAI.com
Get business pricing on office and shipping supplies
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
OpenAI launched 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 its task costs from about $0.05 to $0.40 across effort settings, but its strongest setting scored below several competing models and a cheaper open-weight model outperformed it on both cost and score in the cited comparison.
OpenAI said Terra is available through the API under the name gpt-6.1-terra, as well as in ChatGPT Work and Codex for Plus, Pro, Enterprise and Edu users. The company describes it as suited to communication and computer-based tasks, including its example of processing incoming invoices. OpenAI said the same allowance in ChatGPT Work can cover substantially more tasks than with Sol or Astra.
The listed API price is half that of GPT-5.6 Terra, according to the launch material: $1 per million input tokens and $5 per million output tokens, with cached input at $0.05 per million. These token rates do not, by themselves, show the cost of completing a task. Artificial Analysis’s average task-cost measure accounts for tokens used across its evaluations, including reasoning and cached input.
At maximum effort, Artificial Analysis gave Terra an Intelligence Index score of 48.1 and an average cost of $0.40 per task. Its figures put GPT-6.1 Sol at 51.8 points and $0.72 per task, and Claude Opus 5.5 at 57.6 points and $5.98 per task. The scores and costs summarize the firm’s benchmark evaluations; they do not establish how every user’s workload will perform.
OpenAI also published results from its own tests. It reported a 57.6% score on Agents’ Last Exam for Terra at medium effort, compared with 42.6% for GPT-5.6 Terra at medium effort, and listed costs of $0.29 and $0.84 per task, respectively. On Terminal-Bench 4.0, OpenAI reported that Terra at maximum effort solved 50.3% of tasks, versus 21.5% for the predecessor’s published result. Those are company-reported measurements, distinct from Artificial Analysis’s independent evaluations.
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 Depend on Effort
Terra’s launch highlights a practical issue for organizations choosing models: price per token is not the same as price per completed task. A model with a higher token rate may use fewer tokens to finish work, while asking a model to reason longer can raise costs. Artificial Analysis found Terra’s low and medium settings on its cost-efficiency frontier for the models and settings it tested; the firm’s summary says options above those settings did not offer the same cost-and-score trade-off.
The six-setting results show the trade-off within Terra itself. Artificial Analysis measured scores ranging from 33.5 to 48.1 and task costs from about $0.05 to $0.40. Maximum effort cost 7.4 times as much as low effort for 12 more index points, according to the source material. The lowest-token-price comparison also needs care: Luna is listed at one-tenth Terra’s per-token price, but the reported low-effort Terra task cost was lower because Terra used fewer tokens in that evaluation.
For companies running large volumes of routine work, even small differences in average task cost can affect operating budgets. But benchmark averages are not a promise of savings: actual costs will depend on the task, prompt, output length, settings and usage. More demanding coding or reasoning work may still justify a stronger model, even at a higher price.
Products Worth Considering
Samsung 27 Odyssey OLED G5 (G50SF) QHD QD-OLED Gaming Monitor
Samsung 27-Inch Odyssey QD-OLED G8 4K 240Hz Gaming Monitor
Dell UltraSharp 32 U3226QER 4K Thunderbolt Hub Monitor
As an affiliate, we earn on qualifying purchases.
Where Terra Fits in OpenAI’s Lineup
OpenAI places Terra between Luna and the larger Sol and Astra models. The launch frames it as an option for everyday computer-based work, while acknowledging that its larger models are better suited to the most complex coding tasks. OpenAI said 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.
Artificial Analysis’s maximum-effort comparison also shows that Terra is not the highest-scoring model in the supplied results. Claude Opus 5.5, Kimi K3.1 Pro and Gemini 4 Argon all scored higher. The report also identifies Xiaomi’s open-weight MiMo-V2.6-Pro, at 46.3 points and $0.13 per task, as both cheaper and higher-scoring than Terra’s high and extra-high settings in that comparison.
OpenAI reported other changes, including lower error rates on an internal set of prompts designed to induce factual errors. It said the rate fell from 53.6% to 16.7% at low effort and from 22.3% to 4.3% at maximum effort, comparing Terra with GPT-5.6 Terra. The figures are OpenAI’s internal test results, not an independent finding across all uses.
Benchmark Limits and Open Questions
The available figures do not establish how Terra will perform across all workplace tasks or what an individual customer will pay for a finished job. Artificial Analysis reports averages across its evaluation set, while OpenAI’s figures come from company-selected tests. Results may differ by workload, effort setting and token use.
The source material gives the range across Terra’s six effort settings and specific figures for non-reasoning, low and maximum effort, but does not provide every value in the setting-by-setting table. It reports that the non-reasoning mode cost almost twice as much per task as low effort while scoring lower; the full comparison figures are not included here. Independent speed measurements were also not available in the cited report.
OpenAI said Terra meets the “Critical” threshold in cybersecurity under its Preparedness Framework and retains safeguards developed for GPT-6 Astra. That classification is OpenAI’s own assessment. The launch material does not provide an independent evaluation of that claim.
Watch Real-World Costs and Testing
The next useful evidence will come from independent testing across additional tasks, including speed measurements and more detail on each effort setting. Customers evaluating Terra can compare it with alternatives on their own workflows, recording both task success and the total cost of completed work rather than relying only on per-token rates.
Organizations can also test whether routing routine work to Terra and reserving Sol or Astra for harder tasks suits their needs. OpenAI’s launch announcement does not establish how much any particular customer will save. Actual results will depend on task mix, model settings and usage.
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, with cached input at $0.05 per million. Artificial Analysis measured average task costs of about $0.05 to $0.40 across Terra’s effort settings in its benchmark.
Is Terra OpenAI’s most capable model?
No. OpenAI positions Terra below its larger Sol and Astra models. In Artificial Analysis’s reported maximum-effort comparison, Terra scored below Sol, Astra and several other listed models.
Which Terra effort setting is cheapest?
Artificial Analysis reported the low-effort setting at about $0.055 per task, the lowest cost in the setting figures supplied. The non-reasoning mode was reported to cost almost twice as much per task, despite scoring lower.
Does Terra outperform GPT-5.6 Terra?
OpenAI and Artificial Analysis reported improvements on selected benchmarks. Artificial Analysis said maximum-effort Terra scored six points higher than GPT-5.6 Terra at less than a third of its task cost in that comparison. These results do not show that Terra will be better on every task.
Where can users access GPT-6.1 Terra?
OpenAI says Terra is available through the API as gpt-6.1-terra and in ChatGPT Work and Codex for Plus, Pro, Enterprise and Edu users.
Source: ThorstenMeyerAI.com
