Guides / Language models

GPT-6.1 Terra explained: price, context and where it fits

Updated November 2, 2026. Researched and drafted with AI help; every figure is checked against the linked source.

OpenAI released GPT-6.1 Terra on October 26, 2026. It is the middle model of the GPT-6 family: cheaper than GPT-6.1 Sol, above GPT-6 Luna, and it succeeds GPT-5.6 Terra as OpenAI’s middle model at half the input price or less. This guide covers what it costs, how that compares with the models people are choosing between, and where the price works for you or against you.

The short version

• Price: $1 input / $5 output per million tokens. Cached input: $0.05 per million (5% of the input price).
• Long prompts: above 272K input tokens, the whole request is billed at 2x input and cache rates and 1.5x output, which works out to $2 / $7.50.
• Context: 1,050,000 tokens, with up to 128,000 output tokens. Knowledge cutoff: June 1, 2026.
• API name: gpt-6.1-terra.
• Batch and Flex: 50% off the standard price, so $0.50 / $2.50.
• Cache writes: the model page lists $1.25 per million for 30-minute retention and $2.50 for 12-hour retention.
• In ChatGPT: OpenAI says it is available to Plus, Pro, Enterprise and Edu users in ChatGPT Work and in Codex.

What OpenAI says it is for

OpenAI’s changelog calls GPT-6.1 Terra “a fast, efficient model for visual reasoning”, aimed at visual reasoning, precise computer use and natural writing. In launch coverage, OpenAI describes it as the everyday model for routine office work such as reading what is on screen, navigating apps and drafting replies, and still points people to GPT-6.1 Sol or GPT-6 Astra for in-depth research and complex coding.

OpenAI also published its own test results. These are the maker’s numbers and have not been independently reproduced:

• Terminal-Bench 4.0 (maximum reasoning effort): 50.3% of tasks solved, against 21.5% for GPT-5.6 Terra.
• Speed: OpenAI says Terra completes tasks more than three times faster on average than GPT-6 Astra and GPT-6.1 Sol.

Whether it is good enough for your work is something only your own prompts can show. The GPT-6.1 Terra page collects community votes on how it feels day to day, next to official incidents.

Price next to the alternatives

* Claude Haiku 5.5 prices apply to prompts up to 100K tokens; above that Anthropic lists $0.50 / $2.50.

† Promotional price through December 31, 2026. From January 1, 2027 Google lists $1.50 / $7.50 with $0.15 cache reads.

‡ DeepSeek peak-hour price (weekdays 01:00–04:00 and 06:00–10:00 UTC). At other times it charges half.

° Introductory price through November 30, 2026. From December 1 Google lists $4 / $20 with $0.20 cache reads.

§ Public preview. Mistral cut the price from $1.36 / $4.18 ($0.14 cache reads) to $0.70 / $2.10 ($0.07) when it released the open weights in late October 2026.

What it costs on real workloads

These are the same workloads as our cost per 1,000 requests guide, plus a long-writing case: a chat (2,000 tokens in, 500 out), a coding-agent step (2,000 fresh input tokens, 18,000 read from cache, 1,000 out), a summary (10,000 in, 300 out) and a long write (2,000 in, 8,000 out). Each figure is the cost of 1,000 requests at standard prices.

Where GPT-6.1 Terra fits

Exactly half of GPT-6.1 Sol. Every price is half of Sol’s ($1 vs $2 input, $5 vs $10 output, $0.05 vs $0.10 cache), so every workload costs half: $7.90 against $15.80 per 1,000 agent steps. For a pipeline running one million agent steps a month, that is $7,900 instead of $15,800. If Terra handles the task, the saving needs no other change than the model name.

Less than half of the Terra before it. Against GPT-5.6 Terra, input is half, output is $5 instead of $12 and cache reads are a quarter of the price ($0.05 vs $0.20). On long writing, where output dominates, 1,000 requests cost $42 instead of $100.

Strong on cache-heavy work. The $0.05 cache read is a 95% discount, the same ratio as GPT-6.1 Sol and Claude Sonnet 5.5. That puts Terra’s agent step ($7.90) close to DeepSeek V4 Pro at peak hours ($7.39), even though Terra’s output price is higher.

Not the cheapest for output-heavy work. DeepSeek V4 Pro ($3.96) and Mistral Large 4 ($2.10) charge less per output token, so they beat Terra on long writing ($34.32 and $18.20 against $42.00). Since its late-October price cut, Mistral Large 4 costs less than Terra on every workload here, including agent steps ($4.76 vs $7.90). Gemini 3.8 Flash is cheaper on every workload here while its promotional price lasts; from January 1, 2027 its listed prices double, which would put it above Terra on every column.

Budget models are still in another class on price. GPT-6 Luna and Claude Haiku 5.5 cost about a tenth of Terra on these workloads. For classification, extraction and short replies, start there and move up only if quality is not enough.

Long prompts

Terra accepts 1.05 million tokens, but like other OpenAI models it doubles input (and cache) prices and multiplies output by 1.5 when a prompt goes over 272K tokens. One request with a 300,000-token prompt and a 2,000-token answer, without caching:

At this size Terra’s advantage over Claude Sonnet 5.5 disappears: both cost about $0.62, because Anthropic keeps Sonnet 5.5 at its standard price across the full 1M context. If most of your prompts are very long, compare on your real prompt sizes, and keep in mind that Anthropic says Claude 4.7 and later models produce about 30% more tokens for the same text, so the same document is not the same token count everywhere.

Paying less than list price

• Batch and Flex: OpenAI’s model page lists both at 50% off, which brings Terra to $0.50 / $2.50. Worth it for evaluations, backfills and other jobs that can wait.
• Prompt caching: reads cost $0.05 per million, but writes are billed too: $1.25 per million for 30-minute retention, or $2.50 for the 12-hour retention OpenAI added on October 20, 2026. Caching pays off from the first re-read with 30-minute retention, or from the second with 12-hour retention. Our prompt caching guide works through the maths.
• Watch the 272K line: splitting or trimming a prompt that sits just over 272K tokens halves its input price.

Should you use it?

• Good fit: agents that re-read a large cached context, computer-use and screen-reading tasks (OpenAI’s stated focus), and anything you run today on GPT-6.1 Sol or GPT-5.6 Terra where Terra’s quality holds up in your tests.
• Check alternatives: long writing and output-heavy jobs, where DeepSeek V4 Pro and Mistral Large 4 cost less; very long prompts, where Claude Sonnet 5.5 costs the same; and simple high-volume tasks, where GPT-6 Luna or Claude Haiku 5.5 cost about a tenth.

Compare it head to head with GPT-6.1 Sol, Claude Sonnet 5.5, Gemini 3.8 Flash or DeepSeek V4 Pro, or put your own token counts into the cost calculator. For the rest of OpenAI’s lineup, see GPT-6 vs GPT-5.6, and to show live community ratings on your own site, grab a badge.

Sources

1. OpenAI: Introducing GPT-6.1 Terra
2. GPT-6.1 Terra model page
3. OpenAI developer docs: pricing
4. OpenAI API changelog
5. GPT-5.6 Terra model page
6. GPT-6.1 Sol model page
7. Decrypt: OpenAI launches GPT-6.1 Terra
8. CXO Digital Pulse: OpenAI releases GPT-6.1 Terra
9. Anthropic: Claude API pricing
10. Google: Gemini API pricing
11. xAI docs: Pricing
12. DeepSeek API pricing
13. Mistral docs: Mistral Large 4
14. Mistral: Mistral Large 4 open weights announcement
15. Mistral docs: API pricing