GPT-6.1 Terra: Model ID, Responses API and Migration Checks

The official GPT-6.1 Terra model ID is gpt-6.1-terra, with a dot between 6 and 1. Use it in the model field of a Responses API request. OpenAI describes Terra as a fast, efficient model for visual reasoning, precise computer use, and clear, natural writing.

LLMAPI access is a separate question: GPT-6.1 Terra is absent from the subscription model list on LLMAPI’s plans page. Do not assume that an official OpenAI listing, an OpenAI-compatible client, or a listing for GPT-6.1 Sol establishes Terra subscription access. Check the LLMAPI model-selection documentation and the subscription model endpoint before configuring an LLMAPI request.

Confirm the model ID and official capabilities

The OpenAI model catalog lists these specifications for Terra:

Setting	Official listing
Display name	GPT-6.1 Terra
Request model ID	gpt-6.1-terra
Inputs	Text and images
Output	Text
Reasoning effort	none, low, medium, high, xhigh, max
Context window	1.05M tokens
Maximum output	128K tokens
Listed tools	Functions, Web search, File search, Computer use

Use the official catalog for OpenAI specifications and pricing, not LLMAPI compatibility metadata. For LLMAPI subscription coverage and plan terms, consult the subscription plans page; those are separate from official OpenAI specifications and prices.

Start with a minimal Responses API request

OpenAI’s text-generation guide demonstrates client.responses.create, an input string, and the SDK’s output_text accessor. The following Python example adapts that documented pattern by selecting Terra.

from openai import OpenAI

client = OpenAI()

response = client.responses.create(
    model="gpt-6.1-terra",
    input="Write a one-sentence bedtime story about a unicorn.",
)

print(response.output_text)

This is an illustrative OpenAI SDK request, not a verified LLMAPI account test. It intentionally does not configure an LLMAPI base URL or claim subscription availability.

For response handling, prefer the documented SDK text accessor when you need the combined text. OpenAI warns against assuming that text is always at output[0].content[0].text: the output array can also contain tool calls, reasoning-related items, and other content.

As an initial integration check, keep the prompt simple and inspect the actual response before adding images or tools. Treat a successful text request as a text-request check, not as a substitute for testing the image and tool workflows your application needs.

Choose reasoning effort before comparing models

The OpenAI model guide identifies medium as Terra’s default reasoning effort. It recommends starting there and increasing effort when the quality improvement is worth the additional time and tokens.

Terra supports none, unlike GPT-6 Astra and GPT-6.1 Sol. This matters when moving a configuration between nearby GPT models: do not carry an effort choice across model boundaries without checking the destination model’s supported values.

For migration from GPT-5.6 Terra, OpenAI specifically recommends comparing representative images and workflows at the same reasoning effort. Use that as the baseline for your evaluation, then vary effort separately. Otherwise, a model change and an effort change become difficult to distinguish.

Keep the first request minimal, as above. Before adding an explicit reasoning configuration, check the request schema used by your selected SDK and API route rather than guessing a parameter shape from the list of supported effort values.

Separate visual analysis from computer-use integration

OpenAI gives product-photo matching, technical-drawing comparison, acting on application screens, and drafting or revising text as example Terra tasks. Its guide directs readers to image inputs for visual analysis and computer use for acting on screens.

Plan separate checks for those two workflows:

• Visual analysis: choose representative product photos, drawings, or screenshots and define what a correct answer should contain. These are suggested evaluation cases, not reported test results.
• Computer use: review the computer-use documentation linked from the official model guide before implementing screen actions. Do not treat the text-only example above as a computer-use configuration.
• Existing tool integrations: if your application calls tools through Chat Completions, OpenAI’s Terra migration guidance says to migrate those requests to the Responses API. Review the tool execution and result-handling path as part of that change, rather than changing only the model string.

The official model catalog lists Computer use as a Terra tool capability. That vendor listing should not be interpreted as proof that a particular LLMAPI billing route or account has passed a computer-use test.

Check LLMAPI billing and access separately

LLMAPI distinguishes model names, protocol compatibility, and billing mode. Its documentation says to copy the exact request model ID and verify that the client protocol supports it.

GPT-6.1 Terra is not included in the GPT subscription list shown on the LLMAPI plans page. The list includes gpt-6.1-sol and gpt-5.6-terra; neither is the same request ID as gpt-6.1-terra. Do not substitute either model while describing the result as a Terra test.

The LLMAPI model-selection guide says the subscription list is synchronized with GET /v1/models. Entries in that endpoint use the description “API compatibility identifier for LLMAPI subscription access.” A compatibility label is not an account-specific execution result and should not be presented as an official vendor specification.

Before attempting an LLMAPI integration:

1. Check the exact ID in the current subscription list or the relevant pay-as-you-go channel catalog.
2. Confirm the supported client protocol for that billing route.
3. Use the matching key type: a subscription key for subscription access, or an API billing key with a separate balance for pay-as-you-go access.
4. Verify a minimal request on your account before evaluating images and tools.

LLMAPI’s setup documentation also states that topping up API credit does not increase subscription quota. A balance top-up is therefore not a substitute for checking subscription model coverage.

Migration checklist

Before replacing an existing GPT integration with Terra, review these items independently:

• Exact model: use gpt-6.1-terra, not a display name or the hyphenated spelling gpt-6-1-terra.
• API path: use the Responses API for the starter request; migrate Chat Completions tool requests as directed by OpenAI.
• Reasoning: begin at medium; for GPT-5.6 Terra comparisons, hold effort constant first.
• Output parsing: use output_text where supported instead of assuming the first output item contains text.
• Workload tests: evaluate representative image tasks and screen-action workflows separately.
• Access and billing: distinguish OpenAI’s documented model capabilities from LLMAPI’s exact model coverage, protocol support, and account results.

This keeps three decisions separate: whether Terra fits the task, whether the request follows the documented API pattern, and whether your chosen account and billing route can actually serve it.