🔍 Read the full analysis: Is GPT-6.1 Terra A Better Fit For Your Budget Than OpenAI’s Top Models? on ThorstenMeyerAI.com
Get business pricing on monitors, keyboards and dev gear
- 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, pricing API use at $1 per million input tokens and $5 per million output tokens. Artificial Analysis measured a $0.40 average cost per task and a 48.1 Intelligence Index score at maximum effort, below several leading models; its low and medium settings offered the strongest measured cost-efficiency.
OpenAI released GPT-6.1 Terra on October 26, pitching it as a lower-cost option for routine work, while launch-day testing by Artificial Analysis found it delivers less capability than several leading models. Terra costs $1 per million input tokens and $5 per million output tokens; at maximum effort, the firm measured an average task cost of $0.40 and an Intelligence Index score of 48.1.
Terra is positioned between OpenAI’s smaller GPT-6 Luna and larger GPT-6.1 Sol and Astra. OpenAI lists it in ChatGPT Work, Codex and its API under the identifier gpt-6.1-terra. Its API rates are half those of GPT-5.6 Terra, according to the launch information, and cached input is priced at $0.05 per million tokens.
Artificial Analysis’s index combines ten evaluations and includes the costs incurred while completing them. At maximum effort, it scored Terra 48.1, compared with 51.8 for GPT-6.1 Sol and 52.7 for GPT-6 Astra. The firm’s reported average cost per task was $0.40 for Terra, $0.72 for Sol, and $3.26 for Astra. Claude Opus 5.5 scored 57.6 at $5.98 per task, while other competitors also ranked above Terra.
Terra has six effort settings, and the reported results show why its price per token alone does not settle the budget question. Artificial Analysis measured scores across settings ranging from 33.5 in non-reasoning mode to 48.1 at maximum effort, with task costs from $0.107 to $0.40. Low effort cost $0.055 per task and medium effort was also among the firm’s most cost-efficient settings. The source says low and medium sit on the measured cost-efficiency frontier; settings above them do not.
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
When Lower Task Costs Matter
For organizations handling many routine, repeatable tasks, the relevant budget may be the cost of a completed job rather than the listed price per token. Artificial Analysis’s measurements suggest Terra at low or medium effort can be economical on its test mix. The result is not a guarantee for every company: task length, input format, retries and the amount of reasoning required can change actual spending.
The trade-off is capability. At maximum effort, Terra’s score remained below Sol, Astra and several outside competitors. OpenAI says its own testing shows Terra can handle computer-based work, but also says Sol and Astra remain better choices for the most complex coding tasks. Buyers therefore face a routing decision: use a cheaper model where its performance is sufficient and reserve more capable models for demanding work.
Token rates can also mislead when models produce different amounts of text. The source reports that Luna’s token prices are lower, yet Terra at low effort costs less per task in Artificial Analysis’s evaluation because it generates fewer tokens. Those figures apply to the benchmark tasks, not every user’s workload. A team considering Terra should measure its own completed-task costs and error rates before changing a production budget.
Popular right now
KitchenAid Artisan 5-Qt Tilt-Head Stand Mixer, Cardamom
Ninja Foodi 10-Qt DualZone Smart XL Air Fryer, 6-in-1, DZ550
Shark AI Ultra Robot Vacuum with Self-Empty Base
As an affiliate, we earn on qualifying purchases.
Terra’s Place in OpenAI’s Lineup
OpenAI describes Terra as a “fast, efficient model” for communication and computer-based tasks. Its example is processing incoming invoices and entering their details into accounting software. OpenAI says users on Plus, Pro, Enterprise and Edu can access it through ChatGPT Work and Codex, where the same allowance is expected to cover substantially more tasks than with Sol or Astra. That allowance comparison is OpenAI’s description, not an independently verified promise for every workload.
OpenAI also published its own task results. On Agents’ Last Exam, it reported Terra at medium effort scored 57.6% and cost $0.29 per task, compared with 42.6% and $0.84 for GPT-5.6 Terra at medium effort. OpenAI said Terra at maximum effort cost $1.01 per task on that test, against $8.82 for Claude Opus 5.5, while Opus scored higher: 63.2% to Terra’s 60.0%. These are OpenAI’s measurements and should be kept separate from Artificial Analysis’s index and cost figures.
OpenAI further reported lower error rates on its internal error-inducing prompts and a 50.3% result for Terra on Terminal-Bench 4.0 at maximum effort, compared with a published 21.5% for GPT-5.6 Terra. It said Terra completes tasks more than three times faster than Sol and Astra. Artificial Analysis had not published speed measurements for Terra when the source was checked on October 26, 2026, so that speed comparison remains a company claim.
How Results Translate to Workloads
Artificial Analysis’s task costs are averages across its benchmark evaluations, not quotes for a particular company’s use. The supplied results do not provide the complete per-setting table: some intermediate effort scores, token counts and hallucination rates are missing. That limits how precisely readers can compare all six settings from the available figures.
The benchmark evidence also does not establish that Terra will be cheaper or sufficiently accurate for every workflow. OpenAI’s internal tests and Artificial Analysis’s independent index use different evaluations, so their scores should not be treated as directly interchangeable. Details such as the cost of retries, human review, and performance on specialized business data are not established in the source material.
Artificial Analysis reported that Terra guessed on 38% of questions it could not answer in its hallucination test, down from 88% for GPT-5.6 Terra. This is a result on that specific test, not a general measure of factual reliability. OpenAI’s statement that Terra meets the “Critical” threshold in cybersecurity under its Preparedness Framework is also the company’s own classification.
Test Costs Before Scaling
The immediate next step for teams evaluating Terra is to test it against their actual work: track the cost per successfully completed task, measure error rates and review how often a more capable model is needed. Starting with low or medium effort follows the cost-efficiency pattern reported by Artificial Analysis, but those settings still need validation on a company’s own data and requirements.
Further independent measurements could clarify Terra’s speed, reliability and performance across the six effort levels. For now, the launch-day evidence supports a narrower conclusion: Terra offers lower listed token rates than OpenAI’s larger models, and some benchmark settings appear cost-efficient, but the top models score higher on the reported index. The best budget fit depends on whether Terra completes a reader’s particular tasks accurately enough at its lower cost.
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. Cached input is listed at $0.05 per million tokens. Artificial Analysis measured an average task cost of $0.40 at maximum effort on its index, but actual costs depend on the work being done.
Is Terra cheaper than GPT-6.1 Sol?
In Artificial Analysis’s launch-day comparison, maximum-effort Terra averaged $0.40 per task, versus $0.72 for Sol. Terra also scored lower on the Intelligence Index: 48.1 compared with 51.8. Those are benchmark averages, not guaranteed costs or performance for every application.
Which Terra effort setting is the best value?
Artificial Analysis said low and medium effort sat on its cost-efficiency frontier. Low effort averaged $0.055 per task in the supplied figures. The best setting for a given use depends on whether it meets the required accuracy and task-completion standard.
Does Terra outperform OpenAI’s top models?
Not on Artificial Analysis’s reported Intelligence Index at maximum effort: Terra scored 48.1, below GPT-6.1 Sol at 51.8 and GPT-6 Astra at 52.7. OpenAI’s own results show strengths on particular tasks, but the company also says Sol and Astra are better for its most complex coding work.
Where can people use GPT-6.1 Terra?
OpenAI says Terra is available in ChatGPT Work and Codex for Plus, Pro, Enterprise and Edu users, as well as through its API under the model name gpt-6.1-terra.
Source: ThorstenMeyerAI.com
