By 
OpenAI released GPT-6.1 Terra on Monday at $1 per million input tokens and $5 per million output tokens. GPT-6.1 Terra pricing works out to half of GPT-5.6 Terra on input and 58% less on output, but any request over 272,000 input tokens is billed at higher long-context rates for the whole call.
The model is live for Plus, Pro, Enterprise and Edu users in ChatGPT Work and Codex, and in the API as gpt-6.1-terra, according to OpenAI. Amazon made it generally available on Bedrock the same day.
OpenAI pitches Terra as its fast, efficient tier for computer use, image analysis and writing. Its own example is the dull, high-volume kind: reading incoming invoices and typing the details into accounting software.
What is GPT-6.1 Terra pricing per million tokens?
Standard API rates are $1 for input, $0.05 for cached input and $5 for output, per OpenAI's model page and pricing table. Here is how that lines up against the model it replaces and the tier above it.
| Price per 1M tokens | GPT-6.1 Terra | GPT-5.6 Terra | GPT-6.1 Sol | Terra cut vs GPT-5.6 Terra |
|---|---|---|---|---|
| Input | $1.00 | $2.00 | $2.00 | 50% |
| Cached input | $0.05 | $0.20 | $0.10 | 75% |
| Output | $5.00 | $12.00 | $10.00 | 58% |
| 30 minute cache write | $1.25 | $2.50 | $2.50 | 50% |
A few other rates matter once a workload gets real:
- Cache writes cost $1.25 per million tokens for 30 minute retention, or $2.50 for 12 hours.
- Batch and Flex processing are 50% off, which puts Terra at $0.50 input and $2.50 output.
- Fast mode costs 2x Standard.
- Regional processing for data residency adds 10% where available.
Is GPT-6.1 Terra really 50% cheaper?
Yes, and on two of three meters it is cheaper than that. OpenAI's launch post says Terra's standard prices are 50% lower than GPT-5.6 Terra and GPT-6.1 Sol. AWS phrases it as at least 50% lower.
Against GPT-6.1 Sol, the cut is exactly half on every line. Against GPT-5.6 Terra, the line by line math is more generous:
- Input falls from $2 to $1, a 50% cut.
- Output falls from $12 to $5, a 58% cut.
- Cached input falls from $0.20 to $0.05, a 75% cut.
So jobs that generate long responses, or that reuse the same long prompt prefix, save more than the headline figure.
What is the 272K token catch?
Terra has a 1,050,000 token context window and 128,000 max output tokens. But OpenAI's model page says prompts with more than 272K input tokens are priced at 2x input and cache rates and 1.5x output, for the full request.
That makes the meter jump at the line rather than rise gradually. Our arithmetic, using OpenAI's listed rates and no caching:
- 270,000 input tokens plus 20,000 output tokens costs about $0.37.
- 280,000 input tokens plus the same 20,000 output tokens costs about $0.71.
That is 4% more input for roughly 92% more money. Above the line Terra bills $2 input and $7.50 output, which matches Claude Sonnet 5.5 on input and still undercuts it on output.
Anthropic handles this differently. Its pricing page says Claude 4.6 and later models, except Haiku 5.5, bill the full 1M token window at standard rates, so a 900K token request on Sonnet 5.5 costs the same per token as a short one.
Teams stuffing whole codebases or document sets into one call should chunk, retrieve or cache to stay under 272K.
How does GPT-6.1 Terra compare with Claude on price?
On list price, Terra sits at half of Claude Sonnet 5.5 and a quarter of Claude Opus 5.5, based on Anthropic's pricing page. Anthropic, which Google backed with $2 billion in 2023, now sells Sonnet 5.5 at $2 input and $10 output.
To make that concrete, we priced one sample job: 10,000 requests of 3,000 input tokens and 500 output tokens each, at standard rates with no caching.
| Model | Input / output per 1M | Cost of the sample job |
|---|---|---|
| GPT-6 Luna | $0.10 / $0.50 | $5.50 |
| Claude Haiku 5.5 (prompts up to 100K) | $0.10 / $0.50 | $5.50 |
| GPT-6.1 Terra | $1 / $5 | $55 |
| GPT-6.1 Sol | $2 / $10 | $110 |
| Claude Sonnet 5.5 | $2 / $10 | $110 |
| GPT-5.6 Terra | $2 / $12 | $120 |
| Claude Opus 5.5 | $4 / $20 | $220 |
| GPT-6 Astra | $10 / $50 | $550 |
Two caveats. Different models burn different numbers of tokens on the same task, so list price is not cost per task. OpenAI says that at max effort Terra is over three times as token efficient as Opus 5.5, for a roughly 9x lower cost per task, but that is OpenAI's own measurement.
For simple routing or classification, GPT-6 Luna and Claude Haiku 5.5 remain about ten times cheaper than Terra.
What can GPT-6.1 Terra do, according to OpenAI?
Every number in this section is OpenAI's own and has not been independently verified. OpenAI has been pushing ChatGPT toward acting on the web since it added browsing in 2023, and Terra is aimed squarely at that kind of work.
- Agents' Last Exam: at medium effort, 15 percentage points above GPT-5.6 Terra at about one third the cost, and on par with GPT-6 Astra at one tenth its price per token.
- Terminal-Bench 4.0: 50.3% of tasks solved at max effort, versus GPT-5.6 Terra's published 21.5%.
- Factual errors: on OpenAI's internal error-inducing prompts, the error rate at low effort fell from 53.6% to 16.7%. At max effort it was 4.3%.
- Speed: tasks completed over three times faster than GPT-6 Astra and GPT-6.1 Sol.
OpenAI also says GPT-6.1 Sol and GPT-6 Astra remain the better choice for the most complex coding tasks.
What does the GPT-6.1 Terra system card say about safety?
OpenAI's system card addendum says it is treating Terra as Critical in cybersecurity and High for biological and chemical capability under its Preparedness Framework. Terra runs the same safeguards stack as GPT-6.1 Sol and GPT-6 Astra. OpenAI set up its preparedness team in 2023.
In a simulated deployment across 48,690 internal Codex tasks, Terra drew 22 flags at severity 3 or higher, or 0.045%. That compares with 24 for GPT-6.1 Sol, 26 for GPT-6 Astra, 38 for GPT-6 Sol and 61 for GPT-5.6 Sol.
The launch post leaves out one detail. Relative to GPT-6.1 Sol and GPT-6 Astra, OpenAI observed higher rates of flags in several categories, including concealing uncertainty, circumventing restrictions, exfiltration, destructive action and inbound prompt injection. Teams handing Terra a browser or a shell should keep permissions tight.
What should teams do before switching?
- Rerun cost models with your real input to output mix, since the output cut is deeper than the input cut.
- Cap prompts below 272K input tokens unless the long-context rate is worth it.
- Move back-office jobs such as invoice extraction to Batch or Flex for another 50% off.
- Plan for the Responses API, which OpenAI's model page says is required for tool calling.
- Do not plan on custom training: the model page lists fine-tuning as not supported.
- Run your own evals before trusting vendor benchmarks.
More coverage of the company is on our OpenAI tag page.
GPT-6.1 Terra FAQ
What is GPT-6.1 Terra?
It is OpenAI's new mid-tier model in the GPT-6 series, released October 26, 2026, and tuned for computer use, visual reasoning and writing.
How much does GPT-6.1 Terra cost?
$1 per million input tokens, $0.05 cached input and $5 output at standard rates. Requests over 272K input tokens cost $2 input and $7.50 output.
Who can use GPT-6.1 Terra in ChatGPT?
Plus, Pro, Enterprise and Edu users, in ChatGPT Work and Codex.
How big is the GPT-6.1 Terra context window?
1,050,000 tokens, with up to 128,000 output tokens.
Is GPT-6.1 Terra cheaper than Claude Sonnet 5.5?
On list price, yes. It is half of Sonnet 5.5's $2 input and $10 output for prompts up to 272K tokens.
Is GPT-6.1 Terra on Amazon Bedrock?
Yes. AWS announced general availability on October 26, 2026.
