GPT-6 and GPT-6.1: Model IDs and Responses API Setup

To start an OpenAI Responses API request, set model to the exact model ID. OpenAI’s model list includes gpt-6-astra, gpt-6.1-sol, gpt-6.1-terra, and gpt-6-luna. If you are migrating from gpt-6-sol, check reasoning settings before switching: GPT-6.1 Sol does not support none reasoning effort, while GPT-6 Sol does.

For LLMAPI access, make a separate check of the billing mode, key permissions, and supported protocol. The LLMAPI catalog lists gpt-6.1-sol as a subscription compatibility identifier; that listing is not an OpenAI model specification or a blanket guarantee of Responses API feature parity.

Choose an exact model ID

The following selection guidance comes from OpenAI’s model catalog and GPT-6 guide. It describes official OpenAI models, not LLMAPI subscription entitlements.

Model	Request ID	OpenAI’s suggested use	Supported reasoning effort
GPT-6 Astra	gpt-6-astra	Demanding reasoning, coding, and professional work	low, medium, high, xhigh, max
GPT-6.1 Sol	gpt-6.1-sol	Complex work with a balance of intelligence and cost	low, medium, high, xhigh, max
GPT-6.1 Terra	gpt-6.1-terra	Visual reasoning, computer use, and writing	none, low, medium, high, xhigh, max
GPT-6 Luna	gpt-6-luna	Focused, high-volume work	none, low, medium, high, xhigh, max

Use these descriptions to select candidates, then evaluate representative prompts and outputs for your application. They are not workload-specific performance guarantees. OpenAI lists a 1.05M-token context window and 128K-token maximum output for each of these four models; consult the official model catalog for specifications and current vendor pricing rather than treating those limits as LLMAPI account limits.

Review Sol migration settings

Do not treat gpt-6-sol and gpt-6.1-sol as interchangeable strings. OpenAI explicitly directs existing GPT-6 Sol users to review migration guidance before switching to GPT-6.1 Sol.

In particular, inspect any request configuration that sets reasoning effort to none. That value is supported by the older GPT-6 Sol but not GPT-6.1 Sol or Astra. OpenAI documents medium as the default for GPT-6.1 Sol and GPT-6.1 Terra. Keep model selection and reasoning configuration together in your migration checklist.

Make a non-streaming Responses API request

OpenAI’s text-generation guide shows this Python SDK pattern. The example below keeps the documented Astra model and uses a generic illustrative prompt; it is not a report of observed output or a tested LLMAPI gateway configuration.

from openai import OpenAI

client = OpenAI()
response = client.responses.create(
    model="gpt-6-astra",
    input="Write a one-sentence summary of why meeting agendas are useful.",
)
print(response.output_text)

The request places the selected ID in model and the text prompt in input. Configure credentials for the environment where you run the SDK; use YOUR_API_KEY as a placeholder in shared configuration examples, not an actual credential.

For another official model, select its exact request ID from the table and review its supported reasoning settings. OpenAI’s GPT-6 guide instructs users to get started by setting model in a Responses API request. Treat a model change as a reason to rerun your application’s representative tests, not merely a display-name update.

Read text without assuming a fixed output position

For a text-only display path, the Python example uses response.output_text. OpenAI explains that some official SDKs expose this convenience property to aggregate text outputs into one string.

Do not replace it with an assumption that the answer is always at:

output[0].content[0].text

The Responses API output array can contain multiple items, including tool calls and reasoning-related data. If your application needs those items, inspect and handle the relevant response types rather than treating the aggregated text as the whole response.

A useful integration review therefore separates two questions:

• Does the application only need the generated text for display?
• Does it also need to process tool calls or other response items?

Use the documented text shortcut for the first case, and review the full response handling for the second.

Check LLMAPI access separately from OpenAI specifications

Start with LLMAPI’s model-selection documentation and live subscription model list. The model-selection documentation says the live catalog and your key permissions are authoritative for model IDs, protocols, and features.

The subscription list includes gpt-6.1-sol with the description “API compatibility identifier for LLMAPI subscription access.” Copy the request ID, not the display name. Do not infer subscription access to Astra, Terra, or Luna from their presence in OpenAI’s official catalog.

LLMAPI’s client documentation lists https://llmapi.pro/v1 as its OpenAI-compatible base URL. Keep that client setting separate from the check that your selected model and key support the protocol and features your application needs. The official SDK example above is an OpenAI request pattern, not a gateway feature-parity claim.

Billing mode also matters:

• Monthly subscription: use a subscription key; supported models share the plan allowance. The subscription model list is synchronized with GET /v1/models.
• Pay-as-you-go API: use a separate API billing key and balance. Check the selected channel’s models, supported protocols, and prices.
• Quota: topping up API credit does not increase subscription quota.

For current subscription options and access to channel pricing, use the LLMAPI plans page. Do not apply OpenAI’s vendor prices to an LLMAPI subscription or metered channel.

Before switching your application

Use this checklist for a focused migration review:

1. Choose the destination: distinguish an official OpenAI integration from an LLMAPI client configuration.
2. Copy the exact request ID: for LLMAPI, check the live catalog and your key permissions rather than relying on the official vendor model list alone.
3. Review reasoning effort: remove any assumption that GPT-6.1 Sol accepts none because GPT-6 Sol did.
4. Match billing mode and key: subscription allowance and pay-as-you-go balance are separate.
5. Check protocol and feature support: confirm the selected LLMAPI model and key support the request path you intend to use.
6. Review response parsing: use output_text where appropriate, and do not assume the first output item contains the final text.
7. Evaluate representative requests: compare the results your application actually needs before adopting the new model configuration.