🔍 Read the full analysis: Why GPT-6.1 Terra Could Be Worth Using Even If It Isn’t The Smartest 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 released GPT-6.1 Terra on October 26, 2026, positioning it as a lower-cost model for routine work rather than its most capable option. Artificial Analysis rated it below several leading models, but found its low- and medium-effort settings on the cost-efficiency frontier in its evaluation; results vary by task and benchmark.
OpenAI 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 priced at $0.05 per million tokens. OpenAI says the model 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.
Artificial Analysis’s launch-day evaluation gave Terra a score of 48.1 at maximum effort, at an average cost of $0.40 per task across its benchmark index. In the same comparison, GPT-6.1 Sol scored 51.8 at $0.72 per task, while Claude Opus 5.5 scored 57.6 at $5.98. These are averages from Artificial Analysis’s evaluation tasks, not a fixed price for every real-world request.
The source’s six-setting figures show a wide trade-off: Terra’s scores range from 33.5 to 48.1, while per-task costs range from $0.055 at low effort to $0.40 at maximum. The supplied data says low and medium effort sit on the evaluation’s cost-efficiency frontier. The non-reasoning mode is not the least expensive: it averages $0.107 per task and scores 33.5, compared with $0.055 and 36.2 at low effort.
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
Lower Effort Can Cut Task Costs
The comparison highlights why token prices alone can mislead. A model with a lower price per token does not necessarily cost less to complete a task: the amount of text it generates also matters. In Artificial Analysis’s measurements, Terra at low effort cost less per task than GPT-6 Luna at maximum effort, despite Luna’s lower listed token price. That finding applies to the benchmark’s tested workload mix, not every use case.
For businesses processing many routine requests, such as invoice entry, the practical question is whether a cheaper setting completes work accurately enough to reduce total operating costs. OpenAI cites invoice processing as an example of the high-volume work Terra is meant to handle. Companies would still need to test performance on their own documents and workflows, including the costs of checking and correcting outputs.
Terra’s lower score also sets a boundary on the case for using it. Artificial Analysis placed several models above Terra at maximum effort, and OpenAI says Sol and Astra remain better choices for its most complex coding tasks. The release is thus relevant as a cost-and-capability option, not evidence that Terra is the strongest model for every task.
Popular right now
Vitamix 5200 Professional-Grade Blender, 64 oz. Container, Black
KitchenAid Classic Series 4.5 Quart Tilt-Head Stand Mixer K45SS, Onyx Black
De'Longhi Linea Classic Espresso Machine with Adjustable Milk Frother
As an affiliate, we earn on qualifying purchases.
How Terra Fits OpenAI’s Lineup
OpenAI describes Terra as a model for communication and computer-based work, with adjustable effort settings ranging from non-reasoning to maximum. The company says it completes tasks more than three times faster than Astra and Sol, but Artificial Analysis had not published speed measurements for Terra when the source report checked. The speed comparison is therefore an OpenAI claim, not an independently verified result in the cited data.
OpenAI also reported results from its own evaluations. On Agents’ Last Exam, it said Terra at medium effort scored 57.6% and cost $0.29 per task, compared with $0.84 for GPT-5.6 Terra at the same effort. On Terminal-Bench 4.0, OpenAI reported a 50.3% solve rate at maximum effort, against 21.5% for GPT-5.6 Terra’s published result. Those comparisons use company-provided benchmark figures and should be distinguished from Artificial Analysis’s independent measurements.
In Artificial Analysis’s index, Terra’s maximum-effort score was higher than GPT-6 Luna’s 38.1, but lower than Sol’s 51.8 and Claude Opus 5.5’s 57.6. The report also listed Xiaomi’s open-weight MiMo-V2.6-Pro at 46.3 for $0.13 per task. Rankings and costs depend on the index’s tasks, model settings and pricing data recorded on October 26, 2026.
Real-World Costs Still Need Testing
The reported cost and score figures come from benchmark evaluations, and may not predict performance on a particular company’s workload. The source does not provide the full set of figures for every effort setting: its supplied table ends after the low-effort row begins to describe medium effort. It reports that low and medium sit on the cost-efficiency frontier, but does not include the complete six-setting data needed to independently check every point on the curve.
It is also unclear how Terra performs on specific production tasks, what the total cost will be when human review and error correction are included, and whether the reported speed advantage holds under independent testing. OpenAI’s internal factual-error results and cybersecurity classification are company-reported; the supplied material does not include an independent replication of those claims.
Artificial Analysis’s hallucination test found Terra guessed on 38% of questions it could not answer, down from 88% for GPT-5.6 Terra, according to the source. That result is limited to the test’s unanswered questions and should not be read as a general real-world error rate. Benchmark results and listed prices can also change as models and provider pricing are updated.
Compare Settings on Your Workload
For users and organizations considering Terra, the next step is to compare low and medium effort against higher settings on representative tasks, tracking completion quality, latency and total cost per finished task. A pilot can also establish how much review is required and whether savings in model usage are offset by correction work.
Artificial Analysis may publish further measurements, including speed data; none was available in the cited report at the time of its check. OpenAI’s benchmark claims and published model pricing may also be updated. Until those comparisons are available, Terra’s strongest supported case is as a lower-priced option to test for routine tasks, while more difficult work may call for Sol, Astra or another higher-scoring model.
Key Questions
What is GPT-6.1 Terra?
GPT-6.1 Terra is an OpenAI model released on October 26, 2026, aimed at communication and computer-based work. OpenAI offers it in ChatGPT Work and Codex for eligible plans and through the API.
Is Terra OpenAI’s highest-scoring model?
No. Artificial Analysis scored Terra 48.1 at maximum effort in its Intelligence Index, below GPT-6.1 Sol at 51.8 and Claude Opus 5.5 at 57.6 in the cited comparison.
Why might Terra cost less per task than a model with lower token prices?
Task cost depends on tokens used as well as the price per token. In Artificial Analysis’s benchmark, Terra’s low-effort setting cost less per task than GPT-6 Luna at maximum effort, even though Luna had a lower listed token price. That result is specific to the evaluation.
Which Terra setting is cheapest in the supplied figures?
The provided data identifies low effort as the least expensive listed setting, at an average of $0.055 per task. The source says low and medium effort sit on the cost-efficiency frontier, but the complete six-setting table was not supplied.
Should companies use Terra for every task?
Not on the evidence provided. OpenAI says Sol and Astra remain better for its most complex coding tasks, and independent benchmark scores place some other models above Terra. Organizations should test Terra against their own tasks and include review and correction costs.
Source: ThorstenMeyerAI.com
