GPT-6.1 Terra API Pricing: $1 In, $5 Out, Half of GPT-6.1 Sol

Short answer. GPT-6.1 Terra, which OpenAI released on October 26, 2026, costs $1.00 per million input tokens, $0.05 for cached input and $5.00 for output on OpenAI’s pricing page, read on October 27, 2026. Every one of those prices is half of GPT-6.1 Sol’s. GPT-5.6 Terra costs 2 times as much for input and 2.4 times as much for output.

In plain words. A token is a short piece of text, about three-quarters of a word. GPT-6.1 Terra is a new OpenAI model. Each token costs half what it costs on GPT-6.1 Sol. A sample request costs $0.015, against $0.03 on GPT-6.1 Sol.

OpenAI’s changelog has an entry dated Oct 26 under “October, 2026” that reads “Released GPT-6.1 Terra , a fast, efficient model for visual reasoning, precise computer use, and clear, natural writing.” The model page uses the same sentence as its description and adds what it is meant for: “Use GPT-6.1 Terra to match product photos to catalog entries, interpret technical drawings, act on application screens, or draft and revise text.” This post sets out its prices on each of the pricing page’s tabs, what the cache costs, how it compares with GPT-6.1 Sol and with GPT-5.6 Terra, the model OpenAI’s GPT-6 guide talks about migrating from, and what one request costs on each.

The price, tab by tab

OpenAI’s pricing page has tabs for Standard, Batch, Flex, Fast and Ultrafast. In its Markdown version, read on October 27, 2026, each tab is a table with ten price columns: input, output, cached input, a 30-minute cache write and a 12-hour cache write for short context, then the same five for long context. The page defines them as “Short context: ≤272K input tokens. Long context: >272K input tokens.” The Standard row reads “gpt-6.1-terra | $1.00 | $5.00 | $0.05 | $1.25 | $2.50 | $2.00 | $7.50 | $0.10 | $2.50 | $5.00”.

The Batch and Flex rows are identical: “gpt-6.1-terra | $0.50 | $2.50 | $0.025 | $0.625 | $1.25 | $1.00 | $3.75 | $0.05 | $1.25 | $2.50”. The Fast row reads “gpt-6.1-terra | $2.00 | $10.00 | $0.10 | $2.50 | $5.00 | $4.00 | $15.00 | $0.20 | $5.00 | $10.00”. The model page states the same rules in words: “Batch and Flex prices are 50% lower than Standard.” and “Fast mode prices are 2x Standard.” It also says “Regional processing adds a 10% premium where available.” The Ultrafast table prints two rows, gpt-6-astra and gpt-6.1-sol, and none for GPT-6.1 Terra.

Long prompts cost more. The 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.” On the Standard tab that gives $2.00 for input, $0.10 for cached input and $7.50 for output, the figures the page prints in its long context columns. The rule applies to the whole request, not only to the tokens above 272K.

Fast mode on GPT-6.1 Terra, at $2.00 input and $10.00 output, has the same two figures as GPT-6.1 Sol on Standard. They are two different models; the match is in those two prices only.

What the cache costs

The model page says “Cached input tokens are priced at 5% of the uncached input token rate.” That is the $0.05 on the Standard tab: 5% of $1.00. The cached input price is what OpenAI charges for input tokens it has already seen in an earlier request and kept.

Writing to the cache has its own price. The model page explains the two columns: “Cache-write prices shown use 30-minute retention. Standard cache writes cost $1.25 per million tokens for 30-minute retention and $2.50 for 12-hour retention, before long-context adjustments.” The 12-hour option is recent. OpenAI’s changelog entry of Oct 20 reads “You can now choose 12-hour prompt caching in the API.” and “12-hour cache writes cost 2.5x the input token price.” On GPT-6.1 Terra, $2.50 is 2.5 times the $1.00 input price, and $1.25 is 1.25 times.

GPT-5.6 Terra prices its cached input at $0.20, a tenth of its $2.00 input price, so the cache discount is deeper on the new model: cached input costs $0.20 ÷ $0.05 = 4 times as much on GPT-5.6 Terra.

Against GPT-6.1 Sol and GPT-5.6 Terra

OpenAI’s GPT-6 guide talks about migrating from GPT-5.6 Terra: “When migrating from GPT-5.6 Terra, compare both models on representative images and workflows using the same reasoning effort.” GPT-6.1 Sol is the model listed just above it in the pricing page’s Flagship table. Both rows, from the same Standard table on October 27, 2026: “gpt-6.1-sol | $2.00 | $10.00 | $0.10 | $2.50 | $5.00 | $4.00 | $15.00 | $0.20 | $5.00 | $10.00” and “gpt-5.6-terra | $2.00 | $12.00 | $0.20 | $2.50 | $5.00 | $4.00 | $18.00 | $0.40 | $5.00 | $10.00”.

The last column is a sample request with 10,000 input tokens and 1,000 output tokens on the Standard tab, with no cache.

The sample request, worked out:

• GPT-6.1 Terra: 10,000 × $1.00 ÷ 1,000,000 = $0.01 for input, plus 1,000 × $5.00 ÷ 1,000,000 = $0.005 for output, so $0.015.
• GPT-6.1 Sol: 10,000 × $2.00 ÷ 1,000,000 = $0.02 for input, plus 1,000 × $10.00 ÷ 1,000,000 = $0.01 for output, so $0.03.
• GPT-5.6 Terra: 10,000 × $2.00 ÷ 1,000,000 = $0.02 for input, plus 1,000 × $12.00 ÷ 1,000,000 = $0.012 for output, so $0.032.

A thousand such requests cost $15 on GPT-6.1 Terra, $30 on GPT-6.1 Sol and $32 on GPT-5.6 Terra. Price by price, GPT-6.1 Sol costs 2.00 ÷ 1.00 = 2 times as much for input, 0.10 ÷ 0.05 = 2 times as much for cached input, 2.50 ÷ 1.25 = 2 times as much for a 30-minute cache write and 10.00 ÷ 5.00 = 2 times as much for output. GPT-5.6 Terra costs 2.00 ÷ 1.00 = 2 times as much for input, 0.20 ÷ 0.05 = 4 times as much for cached input, 2.50 ÷ 1.25 = 2 times as much for a 30-minute cache write and 12.00 ÷ 5.00 = 2.4 times as much for output. On the sample request that is 0.03 ÷ 0.015 = 2 times for GPT-6.1 Sol and 0.032 ÷ 0.015 = 2.13 times for GPT-5.6 Terra.

On Batch the sample request costs 10,000 × $0.50 ÷ 1,000,000 + 1,000 × $2.50 ÷ 1,000,000 = $0.0075 on GPT-6.1 Terra, and on Fast mode 10,000 × $2.00 ÷ 1,000,000 + 1,000 × $10.00 ÷ 1,000,000 = $0.03. These figures are prices only: they say nothing about how well each model does a given task.

Context, output and what it takes

The model page gives “1,050,000 context window”, “Maximum input tokens: 922,000” and “128,000 max output tokens”, the same three limits as GPT-5.6 Terra’s page. Its knowledge cutoff is “Jun 01, 2026”, against “Feb 16, 2026” for GPT-5.6 Terra. It lists “Input modalities: text, image”, “Output modalities: text” and “Unsupported modalities: audio, video”. The output is text only, which is why the model belongs in the API pricing table, where every model is charged per token and returns text.

Two lines matter for how it is called. The model page says “Tool calling requires the Responses API.” and the GPT-6 guide adds “If your application calls tools through Chat Completions, migrate those requests to the Responses API.” The endpoint table marks “Batch | v1/batch | Supported” and “Fine-tuning | v1/fine-tuning | Not supported”. The only snapshot listed is gpt-6.1-terra.

When it came out, and what it does not deprecate

The changelog dates the release Oct 26, in the “October, 2026” section. OpenAI’s news feed also lists a post titled “Introducing GPT-6.1 Terra”, dated “Mon, 26 Oct 2026 00:00:00 GMT”, with the summary “Bringing fast and capable computer use to everyone”; the post itself returned an error (HTTP 403) to our script, so nothing from it is quoted here.

OpenAI’s deprecations page, read on October 27, 2026, does not name gpt-6.1-terra, and its newest entry is “2026-10-14: GPT-4 series model deprecations”. The launch deprecates nothing. GPT-5.6 Terra appears on that page only as a recommended replacement for older models, never as a model being shut down, so it is still current, and its model page still prices it at $2 input and $12 output.

This site saw the new model in its daily scan of OpenAI’s pricing page on October 27, 2026, and added it the same day after reading the pages above.

The model has its own page with every figure and its source: GPT-6.1 Terra. The two models in the comparison have theirs: GPT-6.1 Sol and GPT-5.6 Terra. Every OpenAI model with a per-token price is on OpenAI API pricing. Batch, Flex and Fast mode across providers are covered in the second price list.

How we checked

OpenAI’s pricing page in its Markdown version (pricing.md), the model pages of GPT-6.1 Terra, GPT-6.1 Sol and GPT-5.6 Terra, the changelog, the GPT-6 guide and the deprecations page were read by a script on October 27, 2026, as did OpenAI’s news feed. GPT-6.1 Terra first appeared in the daily scan of OpenAI’s pricing page that morning; the scan only flags new models and never writes prices. Each price and line quoted here comes from the saved text of those readings, and a test checks every quote and every figure against it. The ratios and the cost of the sample requests are worked out from those prices, with the sums shown.

Frequently asked questions

How much does GPT-6.1 Terra cost?

OpenAI's pricing page, read on October 27, 2026, lists gpt-6.1-terra at $1.00 per million input tokens, $0.05 for cached input, $1.25 for a 30-minute cache write and $5.00 for output, for prompts up to 272K input tokens. Above that, the whole request costs $2.00 input and $7.50 output.

Does GPT-6.1 Terra have Batch, Flex and Fast mode prices?

Yes: the pricing page prints $0.50 input and $2.50 output for both Batch and Flex, and $2.00 input and $10.00 output for Fast mode, and no Ultrafast row. Its model page puts it as 'Batch and Flex prices are 50% lower than Standard.' and 'Fast mode prices are 2x Standard.'

How does GPT-6.1 Terra compare with GPT-5.6 Terra and GPT-6.1 Sol?

GPT-5.6 Terra costs $2.00 input and $12.00 output, and GPT-6.1 Sol $2.00 input and $10.00 output, per million tokens. A request with 10,000 input and 1,000 output tokens costs $0.015 on GPT-6.1 Terra, $0.03 on GPT-6.1 Sol and $0.032 on GPT-5.6 Terra.