🔍 Read the full analysis: A Closer Look At GPT-6.1 Terra’s Promise Of Affordable AI 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 October 26, 2026, positioning it as a lower-cost model for everyday work. Artificial Analysis measured its average task cost at $0.40 at maximum effort and found low and medium settings on its cost-efficiency frontier; the results are benchmark-specific, and stronger models score higher on some tests.
OpenAI describes Terra as a fast, efficient option for routine work, including reading invoices and entering information into accounting software. The launch report says it is available to Plus, Pro, Enterprise and Edu users in ChatGPT Work and Codex, as well as through the API under the model name gpt-6.1-terra. OpenAI says its subscription allowance can cover substantially more tasks with Terra than with Sol or Astra; that is a company claim, and the source provides no usage totals to quantify it.
Artificial Analysis’s launch-day evaluation gives Terra a 48.1 Intelligence Index score at maximum effort, for an average task cost of $0.40 across its evaluation set. The firm’s index combines ten assessments, including agentic office work, terminal tasks and science. Its figures put GPT-6.1 Sol at 51.8 points and $0.72 per task, while Claude Opus 5.5 scored 57.6 for $5.98. These are averages across the index’s tasks, not a price quote for every real-world job.
Terra has six effort settings, and cost rises with effort. Artificial Analysis measured scores ranging from 33.5 at non-reasoning to 48.1 at maximum effort, with per-task costs from $0.107 to $0.40. Low and medium effort were on the firm’s cost-efficiency frontier; higher settings were not. The report says maximum effort cost 7.4 times as much as low effort for 12 additional index points. Non-reasoning was not the cheapest setting: it cost almost twice as much per task as 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
Why Task Costs Change the Comparison
The launch puts attention on a practical distinction: token price is not the same as task cost. A model with a low rate per million tokens may use more tokens to finish a job, while another may complete it with less output. The source report says GPT-6 Luna is ten times cheaper per token than Terra, yet Terra at low effort costs less per task in Artificial Analysis’s evaluation because it produces fewer tokens. That result depends on the tested tasks and should not be assumed to hold for every workload.
For businesses buying AI services at scale, the relevant question is how much it costs to complete a useful task at an acceptable quality level. Terra’s measured low- and medium-effort results could make it a candidate for high-volume work, while difficult coding or reasoning jobs may justify a more capable, more expensive model. Artificial Analysis also found that Xiaomi’s open-weight MiMo-V2.6-Pro scored 46.3 for $0.13 per task, outperforming Terra’s high and extra-high settings on both measures in that comparison. Terra is therefore one option in a competitive cost-performance field, not the sole low-cost choice.
For individual users, a lower-priced model may stretch usage allowances, but a cheaper run is not automatically a successful one. OpenAI itself says Sol and Astra remain better choices for its most complex coding tasks. Buyers should weigh task completion, error rates, speed and the cost of human review alongside headline model prices.
Products Worth Considering
LG 34GX90SA-W 34-inch Ultragear WQHD OLED Curved Gaming Monitor
Samsung 27-Inch Odyssey QD-OLED G8 4K 240Hz Gaming Monitor
Samsung 27 Odyssey OLED G5 (G50SF) QHD QD-OLED Gaming Monitor
As an affiliate, we earn on qualifying purchases.
How Terra Fits OpenAI’s Lineup
OpenAI places Terra between its smaller Luna model and larger Sol and Astra models. The source report lists Terra’s API price at $1 per million input tokens and $5 per million output tokens, compared with $2 and $12 for GPT-5.6 Terra; cached input is listed at $0.05 per million tokens. OpenAI says the new API rates are 50% lower than those for GPT-5.6 Terra and GPT-6.1 Sol.
OpenAI’s own launch-page charts report gains over GPT-5.6 Terra on selected tests. At medium effort on Agents’ Last Exam, Terra scored 57.6% versus 42.6% for its predecessor, at a reported $0.29 per task versus $0.84. On Terminal-Bench 4.0 at maximum effort, OpenAI reports Terra solved 50.3% of tasks, compared with 21.5% for the predecessor’s published result. These are company-reported benchmark results; they are not equivalent to Artificial Analysis’s independent index and should be read separately.
OpenAI also reports lower error rates on its internal set of prompts designed to induce factual errors: 16.7% at low effort and 4.3% at maximum effort, compared with 53.6% and 22.3% for GPT-5.6 Terra at the corresponding settings. The report says Artificial Analysis’s separate hallucination test found Terra guessed on 38% of questions it could not answer, down from 88% for GPT-5.6 Terra. The tests use different methods, and neither figure alone establishes performance across all uses.
Limits of the Available Comparisons
The published figures do not establish how Terra will perform across every customer’s tasks, or whether the benchmark cost averages will match production bills. Artificial Analysis’s figures describe its own evaluation set, while OpenAI’s charts use company-selected tests; the results are not directly interchangeable. The source report also says Artificial Analysis had not published speed measurements for Terra when its figures were retrieved on October 26, 2026, despite OpenAI’s claim that Terra completes tasks more than three times faster than Astra and Sol.
Other points remain unsettled, including how often users will need to move from low or medium effort to a more expensive setting, how the model performs in specific business workflows, and whether its reported gains hold under independent testing beyond the cited benchmarks. OpenAI’s statement that Terra meets the “Critical” cybersecurity threshold under its Preparedness Framework is also the company’s own classification, not an independent certification described in the source.
What Buyers Should Check Next
Organizations considering Terra can test it on representative tasks and compare cost per successfully completed job, including retries, review and any escalation to Sol or Astra. The source report recommends starting at low or medium effort and routing harder tasks to larger models, but the right mix will depend on a team’s accuracy requirements and workload.
Further independent speed data and evaluations across real-world deployments would help clarify Terra’s trade-offs. For now, the clearest available picture is a lower-priced model with promising results on selected routine tasks, alongside evidence that extra effort raises costs and that stronger alternatives remain ahead on some benchmarks.
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, aimed at routine communication and computer-based work.
How much does GPT-6.1 Terra cost?
The source report lists API rates of $1 per million input tokens and $5 per million output tokens, with cached input at $0.05 per million tokens. Artificial Analysis estimated an average task cost of $0.40 at maximum effort on its evaluation set; actual costs vary by task and setting.
Is Terra cheaper than GPT-6 Luna?
Terra is not cheaper per token: the report says Luna’s token rate is ten times lower. In Artificial Analysis’s tested tasks, however, Terra at low effort cost less per completed task, which the report attributes to lower token use. That benchmark result may not apply to other workloads.
Which Terra setting should users start with?
The source report suggests beginning with low or medium effort for routine work, then using higher effort or a larger model when a task requires more capability. Testing with a team’s own tasks can show whether the added cost improves results enough to justify it.
Does Terra outperform OpenAI’s larger models?
Not across the cited comparisons. Terra scored below Sol and Astra on some measures, and OpenAI says Sol and Astra remain better for its most complex coding tasks. Terra’s case is its lower price and performance on selected routine work, not universal superiority.
Source: ThorstenMeyerAI.com
