🔍 Read the full analysis: GPT-6.1 Terra: OpenAI’s Model For Getting More Done On A Budget on ThorstenMeyerAI.com
Get the little things that make your day 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, positioning it as a lower-cost model for routine work. Artificial Analysis measured a range of performance and per-task costs across six effort settings; its results indicate that low and medium effort offer the strongest cost-efficiency, while harder tasks may call for larger models.
OpenAI released GPT-6.1 Terra on October 26, describing it as a fast, efficient model for everyday communication and computer-based tasks. Its API price is $1 per million input tokens and $5 per million output tokens, half the listed rates for GPT-5.6 Terra, while independent testing found that the cost of completing a task varies substantially with the model’s effort setting.
Terra is available through the API as gpt-6.1-terra and to Plus, Pro, Enterprise and Edu customers in ChatGPT Work and Codex, according to the launch material. OpenAI says the model is designed for high-volume work such as reading invoices and entering information into accounting software. It also says eligible ChatGPT users can complete substantially more tasks with the same allowance than with Sol or Astra; the company did not quantify that claim in the supplied material.
Artificial Analysis, an independent benchmarking firm, published measurements on launch day. It gave Terra an Intelligence Index score of 48.1 at maximum effort, at an average cost of $0.40 per task across the index’s evaluations. GPT-6.1 Sol scored 51.8 at $0.72 per task, while Claude Opus 5.5 scored 57.6 at $5.98. Those figures compare results on the benchmark’s task set, not the price of every real-world job.
Terra has six effort settings, ranging from non-reasoning to maximum. The source material reports that scores rose from about 33 to 48 across the settings, while average task costs ranged from roughly $0.05 to $0.40. Low and medium effort were on Artificial Analysis’s cost-efficiency frontier; settings above them cost more without matching that trade-off. The supplied figures do not include the complete results for every setting.
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
Choosing Effort by Task Cost
For businesses that run large numbers of routine model tasks, cost per completed task may be more useful than token rates alone. A cheaper token price does not automatically mean a cheaper result: models can use different numbers of input, output and reasoning tokens. Artificial Analysis’s measurements found that Terra at low effort cost less per task than the smaller GPT-6 Luna in its evaluation, despite Luna’s lower listed token price.
The trade-off is performance. Terra did not lead the benchmark, and the source says its maximum-effort score remained below several models that scored higher. For users and developers, the results point to routing work by difficulty: use lower settings for routine tasks, then consider a more capable model when accuracy or complex reasoning matters. That is an interpretation of the reported tests, not a guarantee of savings or quality on a particular workload.
OpenAI also reported improvements on its own coding and error tests. On Terminal-Bench 4.0, it said Terra solved 50.3% of tasks at maximum effort, compared with 21.5% for GPT-5.6 Terra’s published result. OpenAI cautioned that Sol and Astra remain better choices for its most complex coding tasks.
Products Worth Considering
Philips Hue Festavia Globe Outdoor String Lights, 32 Bulbs, 91 ft
Bose SoundLink Flex Bluetooth Speaker (2nd Gen), Portable Waterproof Speaker, Black
Philips Hue Large Smart Light Tube, Black – White and Color Ambiance LED Color-Changing Light – 1 Pack
As an affiliate, we earn on qualifying purchases.
Terra Between Luna and Sol
OpenAI places Terra between its smaller Luna model and larger Sol and Astra models. The launch presents Terra as a mid-range option, with lower API rates than GPT-5.6 Terra and GPT-6.1 Sol. Listed prices for Terra are $1 per million input tokens and $5 per million output tokens; cached input is listed at $0.05 per million tokens.
Artificial Analysis’s Intelligence Index combines ten evaluations, including agentic office work, terminal tasks, science and hallucination tests, and tracks costs associated with those evaluations. Its figures are averages across that benchmark and can differ from costs on an individual company’s workload. The firm also reported a 48 score on its Coding Agent Index for Terra, compared with 54 for Sol and 31 for Luna.
In a hallucination test, Artificial Analysis said Terra guessed on 38% of questions it could not answer, down from 88% for GPT-5.6 Terra. OpenAI’s separate internal tests reported lower factual-error rates than its predecessor at low and maximum effort. These are different evaluations, and their results should not be treated as directly interchangeable.
Benchmark Results and Open Questions
The reported costs and scores come from benchmark evaluations, not a guarantee of performance or expense on a customer’s own tasks. Real-world costs will depend on task type, token use, effort setting and other implementation choices. The supplied source does not provide the complete six-setting table, so exact figures for all settings cannot be independently summarized here.
OpenAI’s claims about speed, allowances, computer-use performance, cybersecurity classification and internal error rates are company-reported. Artificial Analysis had not published speed measurements for Terra in the material available on launch day. The company’s statement that Terra meets a cybersecurity threshold is its own classification, not an independent finding in the supplied benchmarking report.
It is also unclear how Terra will perform across customer workloads or how its pricing and availability may change. Benchmark scores cannot by themselves establish whether Terra is the best-value choice for a specific use case.
Testing Terra in Everyday Work
Developers and organizations can test Terra through the API, while eligible subscribers can use it in ChatGPT Work and Codex. The next useful evidence will be independent speed measurements, fuller reporting across all effort settings, and evaluations on real workloads that show both completion quality and total cost.
For now, Artificial Analysis’s results suggest starting with low or medium effort for routine jobs and comparing the expense and accuracy with alternatives. More demanding tasks may justify a higher setting or a larger model, but the best choice will depend on the task and the cost of errors.
Key Questions
When did OpenAI release GPT-6.1 Terra?
OpenAI announced Terra on October 26, 2026. The supplied launch information says it is available through the API and to Plus, Pro, Enterprise and Edu users in ChatGPT Work and Codex.
How much does GPT-6.1 Terra cost?
The listed API price is $1 per million input tokens and $5 per million output tokens, with cached input at $0.05 per million tokens. The cost of a finished task varies with token use and effort setting.
Is Terra OpenAI’s highest-performing model?
No. Artificial Analysis scored Terra at 48.1 at maximum effort, below GPT-6.1 Sol at 51.8 and Claude Opus 5.5 at 57.6 in the cited index. Benchmark results describe performance on that test set, not every use case.
Which Terra effort setting is best value?
Artificial Analysis placed low and medium effort on its cost-efficiency frontier. That finding applies to its benchmark; users should test their own tasks because performance and cost can vary.
What remains unverified about Terra?
Independent speed measurements were not available in the supplied launch-day material. OpenAI’s claims about speed, internal error tests and cybersecurity classification are company-reported, and performance on individual customer workloads remains to be established.
Source: ThorstenMeyerAI.com
