🔍 Read the full analysis: GPT-6.1 Terra: A Budget-Friendly OpenAI Model For Everyday Work on ThorstenMeyerAI.com
Get business pricing on tech for your team
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
OpenAI released GPT-6.1 Terra on October 26, 2026, positioning it as a lower-cost model for everyday work and computer-based tasks. Artificial Analysis measured a $0.40 average task cost at max effort and found low and medium settings on its cost-efficiency frontier; Terra still scores below several more capable models, and some comparisons come from OpenAI’s own tests.
OpenAI released GPT-6.1 Terra on October 26, introducing a model aimed at routine workplace and computer-based tasks at $1 per million input tokens and $5 per million output tokens. Independent benchmark data from Artificial Analysis puts Terra’s average cost at $0.40 per task at its highest reasoning setting, while showing that lower settings can be cheaper for work that does not need as much reasoning.
OpenAI says Terra is available in ChatGPT Work, Codex and the API, where it is identified as gpt-6.1-terra. The company describes it as a fast, efficient model for communication and computer-based work, citing invoice reading and data entry as examples. Its API rates are half those listed for GPT-5.6 Terra: $1 versus $2 per million input tokens, and $5 versus $12 per million output tokens. Cached input is priced at $0.05 per million tokens, according to the launch material.
Artificial Analysis, an independent benchmarking firm, gave Terra an Intelligence Index score of 48.1 at maximum effort, at an average cost of $0.40 per evaluation task. The firm’s index combines ten evaluations, including agentic office work, terminal tasks and hallucination tests. In its comparison, Terra scored below GPT-6.1 Sol at 51.8 and Claude Opus 5.5 at 57.6, but its average task cost was lower: $0.72 for Sol and $5.98 for Opus. These are benchmark results, not a guarantee of the same cost or performance on every customer workload.
Terra’s six effort settings produce a trade-off between score and task cost. Artificial Analysis reported a range from 33.5 points and about $0.11 per task in non-reasoning mode to 48.1 points and $0.40 at max effort. Low effort scored 36.2 and averaged $0.055 per task; the source says low and medium effort sit on the benchmark’s cost-efficiency frontier. The supplied launch report says max effort costs 7.4 times as much as low for 12 additional points. Its table also reports that non-reasoning costs nearly twice as much per task as low effort while scoring lower, underscoring that the cheapest token rate does not necessarily produce the cheapest completed task.
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 Shape Terra’s Value
Terra’s launch puts cost per completed task, rather than the sticker price per token, at the center of its pitch. A model may use fewer tokens to finish a job, so a model with a higher per-token rate can still come out cheaper on a given workload. In Artificial Analysis’s reported figures, GPT-6 Luna costs ten times less per token than Terra, yet Terra at low effort has a lower average task cost in the test data because it uses fewer tokens.
For organizations handling repeated, structured work, that difference could affect how many tasks fit within a budget. OpenAI says Terra’s ChatGPT Work and Codex allowances cover substantially more tasks than with Sol or Astra, but that is a company claim and the launch material does not quantify the allowance comparison. The benchmark results offer a reason for businesses to test Terra on their own workloads, while also showing that a lower-cost model may not be the right choice for every task.
Higher effort is not automatically better value. The benchmark curve shows rising scores alongside rising costs, while Artificial Analysis places low and medium effort on its cost-efficiency frontier. The report also notes that Xiaomi’s open-weight MiMo-V2.6-Pro scored 46.3 for $0.13 per task in the same comparison, outperforming Terra’s higher effort settings on both measures. Terra is therefore one option in a competitive low-cost segment, not the benchmark’s top-performing or lowest-cost model overall.
Products Worth Considering
Sony WH-1000XM4C Wireless Noise-Cancelling Headphones, Black
Apple iPad Air 11-inch (M2), 512GB, Wi-Fi, Space Gray
DJI Osmo Mobile 8 Phone Gimbal with Mic Mini Transmitter, Infinity Black
As an affiliate, we earn on qualifying purchases.
Where Terra Sits in OpenAI’s Lineup
OpenAI positions Terra between GPT-6 Luna, its smaller model, and the larger Sol and Astra models. The company’s launch material presents Terra as a choice for high-volume, everyday work rather than the most demanding tasks. OpenAI itself says Sol and Astra remain better choices for the most complex coding problems, a qualification that limits any reading of Terra as a general replacement for those models.
OpenAI’s own evaluations report gains over GPT-5.6 Terra on several tasks. On Agents’ Last Exam, its medium-effort score was 57.6%, compared with 42.6% for GPT-5.6 Terra, at a reported cost of $0.29 rather than $0.84 per task. At maximum effort on Terminal-Bench 4.0, OpenAI reports that Terra solved 50.3% of tasks, compared with 21.5% for GPT-5.6 Terra. These figures are from OpenAI’s launch-page charts and should be distinguished from Artificial Analysis’s independent index results.
OpenAI also reports lower error rates on an internal set of prompts designed to induce factual errors: 16.7% at low effort and 4.3% at max, compared with 53.6% and 22.3% for GPT-5.6 Terra at the corresponding settings. It says Terra completes tasks more than three times faster than Sol and Astra, but Artificial Analysis had not published speed measurements for Terra when the source report checked on October 26. OpenAI further says Terra meets the “Critical” threshold in cybersecurity under its Preparedness Framework; that classification is the company’s own assessment.
Workload Results Still Need Testing
The benchmark figures describe performance on specific evaluations, not the price or accuracy every user will see. Actual task costs will depend on workload, token use and effort setting; the supplied data does not establish how closely Artificial Analysis’s average maps to a particular company’s invoices, codebase or document flow. The source material also does not provide the full values for every point in Terra’s six-setting curve, so comparisons beyond the listed settings should be treated cautiously.
Several performance and speed claims come from OpenAI’s own tests and have not been independently verified in the material provided. Artificial Analysis had not published Terra speed measurements as of October 26, 2026. The launch information also does not specify the exact number of additional tasks included in ChatGPT Work or Codex allowances, or offer a complete account of how Terra’s cybersecurity classification was assessed. Those details remain open to clarification.
Compare Terra on Real Work
The next practical test is whether Terra’s benchmark cost advantages hold for real users. Companies considering the model can compare low and medium effort against their existing model choice on representative tasks, tracking both completion quality and total cost per finished task. OpenAI’s examples point to invoice processing and other repeatable office workflows, while its own guidance leaves the hardest coding work to Sol or Astra.
Further independent measurements could clarify Terra’s speed, performance across its six effort settings and results on a broader range of workplace tasks. As of the October 26 launch, the available evidence consists of OpenAI’s company-reported evaluations and Artificial Analysis’s published index and cost data; no customer adoption figures or longer-term results are included in the source material.
Key Questions
How much does GPT-6.1 Terra cost?
OpenAI lists API pricing at $1 per million input tokens and $5 per million output tokens, with cached input at $0.05 per million tokens. The launch source says the input and output rates are lower than GPT-5.6 Terra’s $2 and $12 rates.
Is GPT-6.1 Terra OpenAI’s most capable model?
No such claim is supported by the source material. Artificial Analysis scored Terra at 48.1 at max effort, below GPT-6.1 Sol, GPT-6 Astra and Claude Opus 5.5 in the reported comparison. OpenAI says Sol and Astra remain better for its most complex coding tasks.
Which Terra effort setting is the best value?
Artificial Analysis placed low and medium effort on its cost-efficiency frontier, meaning no tested model was both cheaper and better in that comparison. The best setting for an individual task still depends on the quality required and the resulting token use.
Why compare cost per task instead of token prices?
Token prices do not show how many tokens a model uses to finish a job. A model with a higher price per token can have a lower total cost per task if it completes the task with fewer tokens; the reported Luna and Terra comparison illustrates that possibility.
Source: ThorstenMeyerAI.com
