The GPT-6 family from OpenAI is now complete on GetVibe API. GPT-6 Astra arrived first, in early September. GPT-6 Luna and GPT-6 Sol went live on September 23, 2026. All three use the same OpenAI-compatible endpoint and one API key. Only the model value changes.
The three models side by side
| GPT-6 Luna | GPT-6 Sol | GPT-6 Astra | |
|---|---|---|---|
| Model ID | gpt-6-luna | gpt-6-sol | gpt-6-astra |
| Positioning | fast, low-cost tier for high-frequency tasks | flagship reasoning, code and agents | long context, image input, configurable reasoning |
| Context / max output | 272K / 128K | 272K / 128K | 272K / 128K |
| Reasoning levels | minimal to max | minimal to max | minimal to high |
| JSON mode | yes | yes | no |
| Tool calling | yes | yes | yes |
Current prices, pulled live from the catalog:
- GPT-6 Luna: $0.0055 input · $0.0274 output per 1M tokens
- GPT-6 Sol: $0.1096 input · $0.548 output per 1M tokens
- GPT-6 Astra: $0.548 input · $2.7398 output per 1M tokens
"Minimal to max" means the full reasoning_effort ladder: minimal, low, medium, high, xhigh, max. Reasoning tokens are billed as output. The reasoning guide explains how to set the level per request.
Which one to pick
- GPT-6 Luna: classification, extraction, routing, short drafts. Use it anywhere you make many small calls and cost per call matters more than depth.
- GPT-6 Sol: hard coding tasks, multi-step agents, anything where you raise reasoning to
xhighormax. - GPT-6 Astra: a middle option with image input and reasoning up to
high. It has no JSON mode, so for strict JSON output use Luna or Sol, or structured outputs with tool calling.
A practical pattern is to start on Luna and move only the requests that fail your quality check to Sol. Both take the same request body, so switching is a one-string change.
Example request
from openai import OpenAI
client = OpenAI(base_url="https://getvibeapi.com/v1", api_key="sk-pv-...")
resp = client.chat.completions.create(
model="gpt-6-sol",
reasoning_effort="high",
messages=[{"role": "user", "content": "Refactor this function and explain the change: ..."}],
)
print(resp.choices[0].message.content)For Codex CLI and other coding agents, see the Codex CLI setup and the integrations list.
Next steps
Open the model pages for GPT-6 Luna, GPT-6 Sol and GPT-6 Astra for cached-input prices and live latency, or compare the whole catalog on the Models page.