🔍 Read the full analysis: The Lower-Cost Case For GPT-6.1 Terra on ThorstenMeyerAI.com
Get the latest gadgets 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 API use at $1 per million input tokens and $5 per million output tokens. Artificial Analysis found its low- and medium-effort settings on the cost-efficiency frontier in its tests, while higher effort improves scores at greater cost. Results are benchmark-specific, and Terra does not top the measured models.
OpenAI released GPT-6.1 Terra on October 26, positioning it as a lower-cost model for routine work rather than its top-performing system. Its API price is $1 per million input tokens and $5 per million output tokens; independent tests from Artificial Analysis suggest the more useful comparison may be cost per completed task, not token price alone.
Artificial Analysis measured Terra’s performance across six effort settings, from non-reasoning to maximum effort. Its Intelligence Index score ranged from 33.5 at non-reasoning to 48.1 at maximum, while the average cost per task ranged from about $0.055 at low effort to $0.40 at maximum. The firm says low and medium effort fall on its cost-efficiency frontier: in that comparison, no tested model was both less costly per task and higher-scoring. That finding applies to the benchmark and models tested, not every real-world workload.
At maximum effort, Terra scored below several larger or more capable models in Artificial Analysis’s table. Claude Opus 5.5 scored 57.6 at an average $5.98 per task, while GPT-6.1 Sol scored 51.8 at $0.72. Terra’s 48.1 score came at $0.40. Xiaomi’s open-weight MiMo-V2.6-Pro scored 46.3 for $0.13 per task, a lower cost and a slightly lower score than Terra’s maximum setting.
OpenAI lists Terra for ChatGPT Work, Codex and the API, where the model identifier is gpt-6.1-terra. The company describes it as suited to communication and computer-based tasks, including processing invoices. It says the model’s API prices are half those of GPT-5.6 Terra and that eligible ChatGPT plans can complete substantially more tasks under the same allowance. These availability and allowance descriptions come from OpenAI; independent testing of actual customer workloads was not provided in the source material.
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 Change the Comparison
Token prices are easy to compare, but they do not show how many tokens a model uses to finish a job. Artificial Analysis’s measurements suggest Terra can be cheaper per task than a model with a lower per-token price when it produces fewer tokens on the tested work. The firm’s example is GPT-6 Luna: Luna costs less per token, but Terra at low effort costs less per task in its benchmark.
For businesses running large volumes of routine tasks, cost per successful completion may be more useful than the headline rate per million tokens. But price is only one factor: a task that needs stronger reasoning, higher accuracy or less human review may favor a larger model. The source does not establish that Terra will reduce total operating costs across organizations, since integration, retries, oversight and task-specific performance also affect spending.
The results also indicate a possible way to allocate work: start routine requests at low or medium effort, then use a more demanding setting or a larger model when the task calls for it. That is an interpretation of the benchmark curve, not a guarantee that the same routing strategy will perform best in every product or business process.
Products Worth Considering
Bose QuietComfort Headphones (2nd Gen), Wireless Noise-Cancelling Over-Ear Headphones, Eucalyptus Green
Amazon Kindle Scribe Colorsoft 64GB (1st Generation)
Amazon Kindle Paperwhite Signature Edition (12th Generation, 2024), 32 GB, Metallic Black
As an affiliate, we earn on qualifying purchases.
Terra’s Place in OpenAI’s Lineup
OpenAI presents Terra between the smaller GPT-6 Luna and larger Sol and Astra models. The launch material describes Terra as a fast, efficient option for workday tasks, while acknowledging that Sol and Astra remain better choices for the most complex coding problems. The distinction is reflected in Artificial Analysis’s maximum-effort scores: Terra at 48.1, Sol at 51.8 and Astra at 52.7.
OpenAI also published its own launch benchmarks. It reported that Terra at medium effort scored 57.6% on Agents’ Last Exam for computer use, compared with 42.6% for GPT-5.6 Terra, and listed costs of $0.29 and $0.84 per task, respectively. On Terminal-Bench 4.0, OpenAI reported a 50.3% score for Terra at maximum effort, against 21.5% for GPT-5.6 Terra. These are company-reported results, separate from Artificial Analysis’s measurements.
On OpenAI’s internal prompts designed to elicit factual errors, the company reported error rates falling from 53.6% to 16.7% at low effort and from 22.3% to 4.3% at maximum effort, comparing Terra with GPT-5.6 Terra. Artificial Analysis separately reported that Terra guessed on 38% of questions it could not answer in its hallucination test, compared with 88% for GPT-5.6 Terra. The measures use different test descriptions and should not be treated as interchangeable.
Limits of the Published Benchmarks
The available results do not show how Terra performs across the full range of customer workloads, or whether benchmark cost savings translate into lower total costs after retries, review and deployment expenses. Artificial Analysis had not published speed measurements for Terra in the source material, so OpenAI’s claim that it completes tasks more than three times faster than Astra and Sol remains a company statement rather than an independently verified comparison there.
Several details also depend on the specific test and setting. The reported scores and per-task costs are averages over Artificial Analysis’s evaluation tasks; they should not be read as a price quote for an individual request. The source provides only partial data for the six-setting curve, so exact results for every effort level cannot be reported here. It is also unclear how the model’s performance will hold up on new evaluations or in production use.
How Users Can Evaluate Terra
Users can access Terra through the listed ChatGPT Work plans, Codex and the API, according to OpenAI. The next useful evidence will be independent speed results, broader evaluations and comparisons based on real workloads rather than benchmark averages alone.
Organizations considering deployment can test representative tasks at low and medium effort, recording completion quality, token use, retries and human review alongside cost. More demanding requests can then be compared with Terra at higher effort or with Sol and Astra. OpenAI’s launch data supports testing this tiered approach; it does not establish which setting or model will be cheapest for a particular organization.
Key Questions
How much does GPT-6.1 Terra cost?
OpenAI’s listed API price is $1 per million input tokens and $5 per million output tokens. Cached input is listed at $0.05 per million tokens in the source material. Cost per task varies with the work and effort setting.
Does Terra outperform OpenAI’s larger models?
Not in the cited Artificial Analysis maximum-effort comparison. Terra scored 48.1, below GPT-6.1 Sol at 51.8 and GPT-6 Astra at 52.7. OpenAI says Sol and Astra remain preferable for its most complex coding tasks.
Why compare cost per task instead of token price?
A model’s token price does not show how many tokens it needs to complete a task. Artificial Analysis measured both score and average task cost, and found that Terra’s low and medium settings offered a strong cost-performance balance in its evaluation.
Where can users access GPT-6.1 Terra?
OpenAI says Terra is available in ChatGPT Work for Plus, Pro, Enterprise and Edu users, as well as in Codex and through the API.
Source: ThorstenMeyerAI.com
