GPT-6 Astra, Sol, and Luna: Choosing a Model and Using the Responses API

The practical migration is to put the chosen GPT-6 model ID in a Responses API request, then update how your application reads the result. For a non-streaming text response, an official SDK’s output_text convenience property can be useful. Do not replace that with an assumption that text always lives at output[0].content[0].text: the output array can contain other items. For streaming, handle typed events and append text deltas rather than treating each event as a complete response.

The SDK examples below follow OpenAI’s model guidance, text-generation guide, and streaming guide. They illustrate the documented SDK calls; they are not a verification that the same calls work through llmapi.pro’s gateway.

Choose Astra, Sol, or Luna

OpenAI describes the GPT-6 family as Astra, Sol, and Luna and recommends choosing according to the reasoning a task requires, latency, and cost. Its starting guidance is gpt-6-astra for the highest level of capability, gpt-6-sol for strong reasoning on demanding tasks, and gpt-6-luna for efficient, repeatable work at scale. Treat those as selection guidance, not a benchmark or a guarantee for your workload.

One configuration difference matters when changing models: Astra does not support none reasoning effort, while Sol and Luna do. If an existing request sets that value, review it before switching the model to Astra. Test your own prompts and expected output with the model you select rather than assuming the three IDs are interchangeable.

llmapi.pro’s plans page lists all three IDs in its GPT series. That is this site’s subscription catalog, not an official OpenAI specification or proof of Responses API feature parity through the gateway. The plans page says the authoritative list of subscription-available models comes from the gateway’s GET /v1/models; consult it when checking availability instead of relying solely on an article.

Make a non-streaming text request

Here is the Python SDK pattern shown in OpenAI’s text-generation guide, with a different illustrative prompt. Configure the SDK’s credentials for the environment in which you run it; use YOUR_API_KEY as a placeholder if documenting a key, never a live credential.

from openai import OpenAI

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

Change model to gpt-6-sol or gpt-6-luna when evaluating those options. The prompt above is generic and does not demonstrate observed model behavior. OpenAI’s text guide says some official SDKs provide output_text to aggregate text outputs into one string. If your integration also needs tool calls, reasoning-related items, or other response data, inspect the relevant items rather than using aggregated text as a substitute for the whole response.

Handle streaming as typed events

For an SDK streaming request, set stream: true. The following JavaScript combines the request and the event-handling pattern in OpenAI’s streaming guide; its prompt is illustrative:

import { OpenAI } from "openai";

const client = new OpenAI();
const stream = await client.responses.create({
  model: "gpt-6-astra",
  input: [{ role: "user", content: "Write a short greeting." }],
  stream: true,
});

for await (const event of stream) {
  if (event.type === "response.output_text.delta") {
    process.stdout.write(event.delta);
  } else if (event.type === "response.completed") {
    console.log("\nResponse completed.");
  } else if (event.type === "error") {
    console.error(event.message);
  }
}

This example displays text deltas and recognizes completion and error events. It is not a handler for every event type: the streaming guide also lists events for other output and tool activity. Decide which events your application needs before migrating a workflow that does more than display text. In particular, do not treat the first event—or the first output item—as the final answer.

Check the gateway boundary separately

If you use llmapi.pro, its client documentation lists an OpenAI-compatible base URL for supported client setups, and its plans page lists the three GPT-6 IDs. The supplied site documentation does not establish that the gateway implements every Responses API behavior illustrated above. Keep model availability, SDK configuration, and Responses feature compatibility as separate checks. The SDK snippets in this article use the official guide’s default client setup; they do not claim to be tested gateway configuration.