🔍 Read the full analysis: The Case For GPT-6.1 Terra When Price Is Part Of The Prompt 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 its average cost at $0.05 to $0.40 per task across effort settings, with low and medium effort on its cost-efficiency frontier; those results are benchmark measurements, not a guarantee of costs for every user.
OpenAI released GPT-6.1 Terra on October 26, a model positioned between its smaller Luna and larger Sol and Astra models, with API prices of $1 per million input tokens and $5 per million output tokens. Independent testing by Artificial Analysis found that Terra’s average cost per benchmark task ranged from $0.05 to $0.40 across its reasoning settings, making the amount of work completed for the price—not token rates alone—the central question for buyers.
Artificial Analysis scored Terra 48.1 on its Intelligence Index at maximum effort, below higher-scoring models including Claude Opus 5.5, Kimi K3.1 Pro and GPT-6.1 Sol. Its reported average cost was $0.40 per evaluation task, compared with $0.72 for Sol and $5.98 for Opus 5.5. The firm’s figures average costs across its benchmark tasks; they do not predict the bill for any particular customer’s workload.
The setting chosen changes the trade-off. Artificial Analysis measured six effort levels, with scores running from 33.5 at non-reasoning to 48.1 at max and average task costs from $0.107 to $0.40. Its summary says low and medium effort sit on the cost-efficiency frontier: in its tested comparisons, no model was both cheaper and better. At the upper settings, the added score came with a larger bill. The source reports max effort costs 7.4 times as much as low for 12 additional index points.
OpenAI lists Terra for use in ChatGPT Work, Codex and the API, with access for Plus, Pro, Enterprise and Edu users in ChatGPT Work and Codex. The company describes it as a fast, efficient model for communication and computer-based work, citing tasks such as processing invoices. OpenAI says its API prices are 50% lower than those of GPT-5.6 Terra and GPT-6.1 Sol; those are listed price comparisons, separate from measured task costs.
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-Level Costs Matter
For organizations running repeated workflows, cost per completed task can be more useful than the price of a million tokens. A cheaper token rate does not automatically mean a cheaper result if a model generates more text, takes longer routes through a task, or needs repeated attempts. Artificial Analysis reports that Terra’s low-effort setting can cost less per task than Luna in its evaluation, even though Luna’s token price is lower. That finding reflects the benchmark’s measured token use and task mix, not every real-world deployment.
The practical implication is to test models on a company’s own work and track quality, completion rate and total cost. A lower-cost setting may suit routine processing, while complex coding or reasoning may call for a larger model. OpenAI itself says Sol and Astra remain better choices for its most complex coding tasks. Terra’s results do not establish that one model or effort level is best across all workloads.
Artificial Analysis also reported a change in its hallucination test: Terra guessed on 38% of questions it could not answer, compared with 88% for GPT-5.6 Terra. This is a result on that firm’s test, not a general measure of factual reliability. Businesses still need safeguards and review appropriate to the consequences of errors.
Popular right now
KitchenAid Artisan 5-Quart Tilt-Head Stand Mixer, Almond Cream
Vitamix 5200 Professional-Grade Blender, 64 oz. Container, Black
Ninja CREAMi 7-in-1 Ice Cream Maker, NC301
As an affiliate, we earn on qualifying purchases.
Terra Between Luna and Sol
OpenAI presents Terra as a middle option in the GPT-6 lineup, between Luna, the smaller model, and Sol and Astra, the larger models. The launch material emphasizes routine, high-volume work. It also says Terra completes tasks more than three times faster than Astra and Sol, though Artificial Analysis had not published speed measurements for Terra at the time of the source report. The speed comparison is therefore an OpenAI claim, not an independently verified result in that evaluation.
OpenAI’s launch-page charts report improvements on several of its own evaluations. For example, it says Terra at medium effort scored 57.6% on Agents’ Last Exam, 15 percentage points above GPT-5.6 Terra at medium effort. OpenAI’s chart also compares the estimated cost per task for those runs. On Terminal-Bench 4.0, the company reports a 50.3% solve rate for Terra at maximum effort, versus 21.5% for GPT-5.6 Terra’s published result. These are vendor-reported measurements and should be read separately from Artificial Analysis’s independent index.
Artificial Analysis’s evaluation combines ten tests, including agentic office work, terminal tasks, science and hallucination, and records estimated task costs. In its results, Terra’s maximum-effort index score was 48.1, compared with 46.3 for open-weight MiMo-V2.6-Pro, which the source says cost $0.13 per task on the same comparison. The benchmark thus places Terra among several options at different price and performance points, rather than establishing it as a universal leader.
What Benchmarks Cannot Settle
The published figures do not show how Terra will perform across every company’s data, workflows, software integrations or usage patterns. Benchmark task costs are averages, and actual API bills depend on input and output volume, caching, selected effort and the work being done. The source material does not provide a full breakdown of every effort setting’s score, cost and token use, so those values cannot all be compared here.
It is also not clear from the available material how OpenAI’s speed and factual-error claims will compare with independent testing. Artificial Analysis had not published speed measurements when the source report was prepared. Its hallucination results apply to its own test set, while OpenAI’s error rates come from an internal set of error-inducing prompts. Neither test alone establishes real-world error rates for all users.
Terra’s inclusion in OpenAI’s cybersecurity preparedness category is also the company’s own classification. The source says OpenAI classified it as meeting the framework’s Critical threshold and retained safeguards developed for GPT-6 Astra; independent verification or additional assessment details were not provided.
Testing Terra in Real Workflows
The next useful evidence will come from independent evaluations of speed, reliability and task costs, alongside customer experience in production. Buyers can compare Terra with alternatives by running representative tasks at different effort settings, recording whether each task succeeds, and calculating total cost per accepted result rather than relying on token prices alone.
OpenAI has made Terra available through ChatGPT Work, Codex and its API, so comparisons can begin with the announced release. Organizations will need to decide which tasks can tolerate a smaller model’s limitations and which require Sol, Astra or another system. Performance, pricing and availability may change; the figures reported here describe the launch-day information and benchmark results dated October 26, 2026.
Key Questions
What is GPT-6.1 Terra?
GPT-6.1 Terra is a new OpenAI model positioned between Luna and the larger Sol and Astra models. OpenAI describes it as intended for communication and computer-based tasks.
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, according to the source material. Artificial Analysis separately measured average task costs from $0.05 to $0.40 across effort settings in its benchmark; actual costs vary by workload.
Which Terra effort setting is the best value?
Artificial Analysis placed low and medium effort on its cost-efficiency frontier in the tested comparisons. The most suitable setting depends on the task, and the benchmark does not establish a universal best choice.
Does Terra outperform OpenAI’s larger models?
Not across the reported comparisons. Artificial Analysis scored Terra at 48.1 at maximum effort, below GPT-6.1 Sol at 51.8 and GPT-6 Astra at 52.7 on its Intelligence Index. OpenAI says Sol and Astra remain better for its most complex coding tasks.
Where can users access Terra?
The source says Terra is available in ChatGPT Work, Codex and the API. OpenAI lists Plus, Pro, Enterprise and Edu users for ChatGPT Work and Codex access.
Source: ThorstenMeyerAI.com
