🔍 Read the full analysis: When GPT-6.1 Terra Could Be A Better Fit Than A Smarter Model on ThorstenMeyerAI.com
Get hardware and tech 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, positioning it as a lower-cost model for routine work rather than its most capable option. Artificial Analysis found that Terra at low and medium effort sits on its measured cost-efficiency frontier, while higher effort costs more for gains in benchmark score. The results are benchmark-specific; complex coding and other demanding tasks may still suit larger models.
OpenAI released GPT-6.1 Terra on October 26, describing it as a fast, efficient model for everyday communication and computer-based work. Its API pricing is $1 per million input tokens and $5 per million output tokens, half the listed rates for GPT-5.6 Terra; independent tests suggest that Terra at low or medium effort can be a better fit for routine tasks than a more capable, more expensive model.
The main distinction is not simply price per token, but the estimated cost of completing a benchmark task. Artificial Analysis reported an Intelligence Index score of 48.1 for Terra at maximum effort, with an average cost of $0.40 per task across its evaluation set. 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 describe performance and average costs on the firm’s tests, not guaranteed costs or results for a particular user’s workload.
Artificial Analysis measured six Terra effort settings, from non-reasoning to maximum. Its data put low effort at $0.055 per task and medium effort at a higher cost, with both settings on the firm’s cost-efficiency frontier: among tested models, no alternative was both cheaper and higher-scoring at those points. Maximum effort reached 48.1, 11.9 points above low effort, but cost about 7.3 times as much. The source’s rounded summary describes the score gain as 12 points and the cost multiple as 7.4.
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. The company says its allowances cover substantially more tasks with Terra than with Sol or Astra, but that is OpenAI’s description of its product limits, not an independently verified usage guarantee.
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
Task Costs Change Model Choice
The launch highlights why token rates alone can mislead organizations comparing AI models. A model that charges less per token may produce longer responses or use more reasoning tokens, raising the price of a completed job. Artificial Analysis found that GPT-6 Luna costs less per token than Terra, yet Terra at low effort had a lower measured cost per task in this evaluation. The result depends on the test’s workload and token use; it does not establish that Terra is cheaper for every application.
For companies handling large volumes of standard work, such as invoice processing or routine computer actions, a modest score at a low cost may be more useful than the highest available benchmark score. Teams can reserve Sol, Astra or another stronger model for tasks where accuracy or complexity calls for it. For individual users, lower task costs may stretch usage allowances, but Terra’s results do not remove the need to check consequential outputs.
The trade-off is visible in the benchmark curve: higher effort raises cost as well as score. Choosing maximum effort for every request may spend more without a proportionate benefit for routine jobs. Conversely, the lowest-cost setting may not be suitable where errors carry serious consequences. The practical question is whether the selected setting completes a specific task reliably enough at an acceptable cost.
Products Worth Considering
Logitech MX Master 4 for Mac, Ergonomic Bluetooth Mouse with Haptic Feedback, Ultra-Fast Scrolling, USB-C Charging, for MacOS, iPadOS, Windo
Logitech MX Mechanical Full-Size Wireless Mechanical Keyboard, Tactile Quiet Switches, Backlit Keys, Bluetooth, USB-C, Multi-Device, Graphit
LG UltraGear 27-Inch QHD OLED Gaming Monitor, 240Hz, 0.03ms, G-SYNC Compatible, FreeSync Premium Pro, HDMI 2.1
As an affiliate, we earn on qualifying purchases.
Terra Between Luna and Sol
OpenAI places Terra between its smaller Luna model and its larger Sol and Astra models. The launch-page description calls Terra “a fast, efficient model” for communication and computer-based tasks that fill the workday. OpenAI’s example is reading incoming invoices and entering their details into accounting software. These are the company’s intended uses, not independent evidence that the model performs every such task accurately.
OpenAI’s own launch-page charts report additional comparisons. On Agents’ Last Exam, it says Terra at medium effort scored 57.6% at $0.29 per task, compared with 42.6% and $0.84 for GPT-5.6 Terra at medium effort. OpenAI says Terra’s result matched GPT-6 Astra at medium effort, whose measured cost was $4.82 per task. On Terminal-Bench 4.0, the company reports Terra solved 50.3% of tasks at maximum effort, against a previously published 21.5% for GPT-5.6 Terra. These are OpenAI-reported comparisons and should be read separately from Artificial Analysis’s independent index.
Artificial Analysis’s index combines 10 evaluations, spanning areas such as office work, terminal tasks, science and hallucination, and includes model spending on the evaluated tasks. Its launch-day assessment also shows Terra is not the highest-scoring option: Claude Opus 5.5, Kimi K3.1 Pro and Gemini 4 Argon scored above it in the listed results. Xiaomi’s open-weight MiMo-V2.6-Pro scored 46.3 for $0.13 per task, below Terra’s maximum-effort score but at a lower measured cost. Comparisons depend on each model’s tested setting and the index’s methodology.
Benchmark Results Have Limits
The published figures do not establish how Terra will perform across every company’s data, software or task mix. Artificial Analysis’s task costs are averages across its evaluation set, while actual charges will depend on prompt length, output, cached input, reasoning use and workload. The available source does not provide a complete breakdown of real-world costs for customer deployments.
Other performance claims remain tied to their source. OpenAI reports that Terra completes tasks more than three times faster than Astra and Sol, but Artificial Analysis had not published speed measurements when the source material was prepared. OpenAI also says Terra meets the Critical threshold in its cybersecurity preparedness framework; that classification is OpenAI’s own assessment, not an independent certification cited here. The material does not establish how either claim translates to every deployment.
Benchmark scores also cannot settle whether a model is dependable enough for high-impact decisions. Artificial Analysis reported lower guessing on questions its test marked unanswerable than for GPT-5.6 Terra, but the result applies to that test. Real-world error rates, availability and customer-specific performance remain unclear from the launch-day information.
Test Settings Against Real Work
The immediate next step for prospective users is to compare Terra’s settings on their own representative tasks, recording both successful completion rates and total cost per finished task. A low or medium setting may be a starting point for routine, high-volume work; cases involving complex coding or difficult reasoning may warrant testing a larger model. The best setting cannot be determined from token rates or a single aggregate score alone.
Further evidence may come from independent speed testing, additional evaluations and customer deployments. Until those results are available, Terra’s launch-day comparisons offer a basis for trial, not a guarantee of savings or quality. OpenAI has not provided, in the cited material, a schedule for further benchmark releases or a quantified commitment about how much more work customers will complete under their specific allowances.
Key Questions
What is GPT-6.1 Terra?
GPT-6.1 Terra is an OpenAI model launched October 26, 2026, positioned for routine communication and computer-based tasks. OpenAI makes it available through the API, ChatGPT Work and Codex for specified subscription tiers.
Why might Terra be a better fit than a smarter model?
On Artificial Analysis’s evaluation, low and medium effort offered a balance of score and average task cost that the firm placed on its cost-efficiency frontier. For routine, high-volume work, that may matter more than reaching the top benchmark score. Results for a particular workload may differ.
How much does GPT-6.1 Terra cost?
OpenAI lists API prices of $1 per million input tokens and $5 per million output tokens, with cached input at $0.05 per million tokens in the source material. Artificial Analysis measured average task costs separately; those are benchmark averages, not a fixed price for every task.
Is Terra OpenAI’s most capable model?
No. The cited independent index scores several models above Terra at maximum effort, and OpenAI says Sol and Astra remain better choices for the most complex coding tasks. Terra is presented as a lower-cost option for a different range of work.
Do the benchmark results guarantee lower costs?
No. They report results under specific tests and settings. Actual cost and quality depend on the task, token use, reasoning setting and how a system handles mistakes, so organizations would need to test their own workflows.
Source: ThorstenMeyerAI.com
