🔍 Read the full analysis: GPT-6.1 Terra Could Be A Smart Way To Keep AI Costs In Check on ThorstenMeyerAI.com
Get everyday essentials 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, pricing its API at $1 per million input tokens and $5 per million output tokens. Artificial Analysis measured lower task costs at low and medium effort, but Terra scores below several higher-performing models, and the benchmark does not establish how it will perform across customers’ workloads.
OpenAI released GPT-6.1 Terra on October 26, presenting it as a lower-cost model for routine work such as processing invoices, while an evaluation by Artificial Analysis found that its low- and medium-effort settings delivered comparatively low costs per benchmark task. Terra’s API price is $1 per million input tokens and $5 per million output tokens; its performance and cost rise with the selected effort level, and it does not top the cited model rankings.
OpenAI says Terra 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. Its listed API prices are half those of GPT-5.6 Terra, according to the launch information: $1 per million input tokens and $5 per million output tokens, with cached input priced at $0.05 per million tokens. OpenAI also says the same allowance in ChatGPT Work can cover substantially more tasks than with Sol or Astra, though the source does not provide a general usage ratio.
Artificial Analysis reported a 48.1 Intelligence Index score for Terra at maximum effort and an average cost of $0.40 per task across its index evaluations. 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 figures describe that benchmark’s tasks and cost method; they are not a promise of what an individual company or user will pay for a completed job.
The effort setting makes a material difference. Artificial Analysis measured Terra’s six settings at scores ranging from 33.5 to 48.1 and average task costs from $0.107 to $0.40. Low effort scored 36.2 at $0.055 per task; the source says low and medium sit on its cost-efficiency frontier, meaning no model it tested was both cheaper and higher-scoring at those points. The provided source excerpt does not include the full results for every setting, so the precise medium-effort figures are not stated here.
OpenAI’s own launch-page results describe other task-specific comparisons. On Agents’ Last Exam, it reports Terra at medium effort scored 57.6% for $0.29 per task, compared with GPT-5.6 Terra’s 42.6% and $0.84. On Terminal-Bench 4.0, OpenAI reports a 50.3% solve rate for Terra at maximum effort, against 21.5% for GPT-5.6 Terra. These are company-reported measurements, separate from Artificial Analysis’s evaluation. OpenAI also says Terra is over three times faster than Sol and Astra, but Artificial Analysis had not published speed measurements for Terra at the time covered by the source.
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 Prices
Terra’s launch puts a practical budgeting question ahead of a simple price comparison: how much does a completed task cost, rather than how much each token costs? Token prices alone can mislead because models may use different numbers of input, output and reasoning tokens to do similar work. Artificial Analysis’s figures suggest that a cheaper-per-token model is not automatically cheaper for a task: it reports GPT-6 Luna at $0.07 per task at maximum effort, while Terra at low effort averages $0.055, despite Luna’s lower listed token prices.
That distinction matters for organizations running large volumes of repetitive work, such as invoice handling, where small per-task differences can affect operating budgets. It also affects how users allocate access: a lower-cost model may be suitable for routine requests, while difficult coding or analysis could require more effort or a stronger model. The benchmark supports testing different settings; it does not establish that Terra will be the least expensive choice for every workflow.
There are trade-offs in the reported results. At maximum effort, Terra’s $0.40 average task cost is below Sol’s and far below Opus 5.5’s in the cited comparison, but its score is also lower. Artificial Analysis lists other lower-cost options too, including open-weight MiMo-V2.6-Pro, which scored 46.3 at $0.13 per task. The commercial decision depends on task accuracy, deployment needs, and the cost of errors—not the benchmark price alone.
Products Worth Considering
Samsung 27 Odyssey OLED G5 (G50SF) QHD QD-OLED Gaming Monitor
LG UltraGear 32GX850A-B 32-Inch 4K UHD Glossy OLED Gaming Monitor
Samsung 32-Inch Odyssey OLED G8 (G80SH) 4K Gaming Monitor
As an affiliate, we earn on qualifying purchases.
Terra Between Luna and Sol
OpenAI describes Terra as a model for communication and computer-based tasks that fill a workday, placing it between the smaller Luna and larger Sol and Astra models. Its launch example is reading incoming invoices and entering their details into accounting software. OpenAI says Terra retains safeguards developed for GPT-6 Astra and meets the Critical threshold in cybersecurity under its Preparedness Framework; those are OpenAI’s own classifications, not an independent assessment in the supplied material.
Artificial Analysis’s Intelligence Index combines ten evaluations, including agentic office work, terminal tasks, science and hallucination testing, and estimates cost across those tasks. Its October 26 model-page figures put Terra below several higher-scoring systems, including Opus 5.5, Kimi K3.1 Pro, Gemini 4 Argon, Astra and Sol. The evaluator also reports Terra guessed on 38% of questions it could not answer in its hallucination test, compared with 88% for GPT-5.6 Terra. That result is specific to the test and should not be read as a general error rate.
Where Benchmark Results Fall Short
The available results do not show how Terra will perform across customers’ real workloads, which can differ from benchmark tasks in data, workflow design and acceptable error rates. The cited cost per task is an evaluation average; actual spending will depend on token use, task length, retries, caching and the effort setting. The source also does not provide a complete six-setting table, leaving some reported curve details unavailable.
Several performance claims come from OpenAI’s launch-page measurements, while the Intelligence Index figures come from Artificial Analysis. Those results use different tests and should not be treated as directly interchangeable. Artificial Analysis had not published speed measurements for Terra in the source material, so OpenAI’s claim that it is over three times faster than Sol and Astra cannot be independently checked against that evaluator there. It is also unclear how broadly OpenAI’s statement about substantially more tasks under the same ChatGPT Work allowance applies across plans and usage patterns.
Testing Terra on Real Workloads
The next useful evidence will come from independent testing across a wider set of tasks, including speed measurements and comparisons at each effort level. Businesses considering Terra can compare its cost per successfully completed task with alternatives using their own representative work, while recording accuracy, retries and human review time. Low or medium effort may be a starting point for routine tasks, but the benchmark alone does not identify the right setting for every use case.
OpenAI’s launch makes Terra available through ChatGPT Work, Codex and the API. The source provides no announced schedule for further evaluations or model changes, so any updates to independent scores, availability terms or pricing remain unconfirmed.
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 source says these prices are half those of GPT-5.6 Terra.
Is Terra the strongest model in the cited comparison?
No. Artificial Analysis gave Terra a 48.1 Intelligence Index score at maximum effort, below several models in its comparison, including Claude Opus 5.5 at 57.6 and GPT-6.1 Sol at 51.8.
Which Terra effort setting is cheapest per task?
Artificial Analysis measured low effort at $0.055 per task on average, the lowest figure listed in the supplied six-setting results. That average applies to the evaluator’s tasks, not every user’s workload.
Where can people use GPT-6.1 Terra?
According to the launch information, Terra is available to Plus, Pro, Enterprise and Edu users in ChatGPT Work and Codex, and through the API as gpt-6.1-terra.
Are Terra’s speed claims independently verified in the cited material?
Not in the source material. OpenAI says Terra is over three times faster than Sol and Astra, but Artificial Analysis had not published speed measurements for Terra when the cited evaluation was checked.
Source: ThorstenMeyerAI.com
