OpenAI Visual reasoning and clear writing
GPT-6.1 Terra
Turn screenshots, reference photos, and source notes into clear decisions your team can review.
GPT-6.1 Terra takes text and images and returns text. Use it when the question is specific and the answer needs to point back to visible evidence, whether you are matching a product photo to catalog entries, triaging an interface screenshot, or turning scattered notes into a handoff.
- Text and image input
- Text output
- 1.05M-token catalog context
The example on this page uses Chat Completions. OpenAI documents computer use through its Responses API. The workflows below use images and text to produce a written result.
AI Chat
Send a prompt and see the response here.
gpt-6.1-terraStart a conversation with this model and ask follow-up questions.
USD /1M tokens
Pricing
| Group | Tier | Input | Output | Cache Read | Cache Write |
|---|---|---|---|---|---|
| default | input≤272k | $1 | $5 | $0.05 | $1.25 |
| default | input>272k | $2 | $7.5 | $0.01 | $2.5 |
Public list prices are shown in USD. Final charges may vary by account group and usage tier.
Overview
What is GPT-6.1 Terra?
Terra is a practical starting point for tasks where the evidence is already in front of you. Give it the image or notes, name the decision, and ask for a short answer that cites what it observed. Make uncertainty part of the response so that close matches, unclear controls, and missing facts get sent to a reviewer.
Use a representative set of easy and borderline cases before choosing a default model. Compare the rate of accepted answers and the time a reviewer spends correcting them. A useful match or handoff is easier to evaluate than an open-ended request to analyze everything.
Separate close matches
Put the reference photo and a small set of labeled catalog candidates in one comparison image. Ask what visible detail distinguishes the candidates.
Describe the state you can see
Use a labeled screenshot and the expected behavior to draft an issue a tester can verify.
Make a handoff actionable
Ask for an explicit decision, next action, and unresolved questions using only the supplied notes.
From brief to deliverable
Put GPT-6.1 Terra to work
Catalog and operations teams
Match a product photo to catalog entries
Narrow a photo to a small set of candidate SKUs. Ask Terra to explain which visible detail supports the match and flag lookalikes for review.
- Tools you need
- A catalog shortlist, a visual review queue and a human for ambiguous matches.
- Keep in mind
- Use visible features such as a seam or control placement. An angle that hides the distinguishing feature is a reason to leave the match unresolved.
View a task brief
In this comparison image, find the catalog tile that best matches the reference. Return the SKU label and the visible details that support it. If the deciding detail is hidden or shared by two candidates, flag the match for review and suggest another photo angle.
Product support and QA
Draft a testable issue from a screenshot
Give Terra the current screen and expected layout. Ask it to locate the visible mismatch, draft a concise issue and separate observations from hypotheses.
- Tools you need
- A screen capture, issue tracker and a reviewer who can reproduce the problem.
- Keep in mind
- A screenshot may show a clipped button without revealing the cause. Verify proposed steps in the actual application before filing.
View a task brief
Read the attached screenshot and expected behavior. Identify the affected area, write observed and expected behavior, then suggest the shortest steps a tester can verify. Label any step that requires confirmation and do not invent a root cause.
Product and engineering teams
Write a handoff from fragmented notes
Group supplied notes into what was decided, who needs to act and which questions still need an answer.
- Tools you need
- Source notes and a reviewer who knows the decision history.
- Keep in mind
- A confident sentence can still merge incompatible notes. Keep conflicting statements visible and leave unconfirmed owners or dates open.
View a task brief
Convert these notes into a handoff with decisions, next actions and open questions. Use the original owner and date only when supplied. If the sources disagree, record both versions and identify what must be confirmed.
Capabilities in context
Key features
| Feature | What you get | Why it matters |
|---|---|---|
| Image understanding | Text and image in, text out | Compare products or review screenshots from a labeled image. The Playground accepts one new image per message. |
| Context window | 1.05M tokens in the TokenHot catalog | Budget long requests using the pricing table. Its tiers are based on input length, while context describes how much information the model can consider. |
| Request protocol | TokenHot Chat Completions example | Start with the current API example for model selection and payload shape. Check other endpoints and tool support separately. |
The TokenHot catalog provides the current route and pricing. OpenAI's model card describes provider-native capabilities and tool availability by endpoint. OpenAI model card · OpenAI model selection
Audience & fit
Where GPT-6.1 Terra fits
Catalog operators
Review photos and product candidates in a repeatable matching workflow.
Product support and QA
Turn visual defects into short, evidence-linked issue drafts.
Product and engineering leads
Turn notes into decisions and next steps without losing uncertainty.
Cost context
Plan the cost of a useful result
Use the active context tier
The current TokenHot table lists one tier for requests with up to 272K input tokens and another above 272K. Check the applicable tier and account group before sending a long history or reference set.
Track review work
Compare accepted matches and reports as well as request cost. A cheaper response can cost more to correct.
Keep the answer focused
Set a compact response format so that a classification or handoff uses only the output it needs.
Use the current pricing table for available rates and account groups. Model IDs and prices come from TokenHot’s catalog; provider pricing and subscriptions are separate.
API
Code examples
curl --location --request POST 'https://api.tokenhot.ai/v1/chat/completions' \
--header 'Authorization: Bearer <token>' \
--header 'Content-Type: application/json' \
--data-raw '{
"model": "gpt-6.1-terra",
"messages": [
{
"role": "user",
"content": "What model are you"
}
]
}'FAQ
Frequently asked questions
What is GPT-6.1 Terra best suited for?
GPT-6.1 Terra is suited to the input, output, and capability types listed on this page. Test production workloads before choosing it for a critical system.
How is GPT-6.1 Terra priced?
Pricing is shown in the pricing section above. Actual cost depends on usage volume, account group, and the request payload.
How can I call GPT-6.1 Terra?
Use the model ID shown above with one of the supported API protocols. Code examples are provided when usage data is available.
What is the TokenHot model ID for GPT-6.1 Terra?
Use gpt-6.1-terra with the TokenHot request example on this page.
Can GPT-6.1 Terra interpret screenshots?
Yes. Attach one screenshot in the Playground and ask for visible evidence. For API requests, use the image payload documented for your endpoint.
Does Terra operate a browser on this route?
The Chat Completions example sends a prompt and returns text. OpenAI documents computer use separately through its Responses API. Confirm that workflow with TokenHot before integrating it.
When should I choose Terra or Sol?
Start with Terra for bounded visual comparisons and clear writing. Include Sol in the evaluation if deeper code or multi-step reasoning materially affects acceptance.
How is GPT-6.1 Terra billed?
Check the current TokenHot pricing table for input, output, cache and context tiers that apply to your account.
Get started
Build with GPT-6.1 Terra
API