GPT-6 API — Astra, Sol and Luna pricing, model IDs and how to call them

GPT-6 is OpenAI's current model family: GPT-6 Astra, released September 3, 2026, and GPT-6 Sol and GPT-6 Luna, released September 22, 2026. OpenAI lists them at $10.00 / $50.00, $2.00 / $10.00 and $0.10 / $0.50 per 1M input / output tokens. On Kunavo all three are live at $4.00 / $20.00, $0.80 / $4.00 and $0.04 / $0.20 — 60% under OpenAI's list — as gpt-6-astra, gpt-6-sol and gpt-6-luna, with a 1,050,000-token context window and up to 128,000 output tokens each.

GPT-6 API pricing

OpenAI's published rates next to Kunavo's, per 1M tokens. Cached input is a tenth of the input rate and cache writes are 1.25 times it, on OpenAI and on Kunavo alike:

A prompt over 272K input tokens bills the whole request at 2x input and cache rates and 1.5x output — OpenAI's tier, applied the same way on Kunavo. Reasoning tokens bill as output on every GPT-6 model. OpenAI's launch note puts Sol and Luna at "50% lower API prices" than GPT-5.6's promotional pricing; on Kunavo, GPT-6 Sol costs $0.80 / $4.00 against GPT-5.6 Sol's $2.00 / $12.00. Every GPT rate is in the GPT API pricing guide.

Which GPT-6 model to call

• Coding agents and multi-step tool use — gpt-6-sol, at a fifth of Astra's price. Start here and move up only where it falls short.
• The hardest long-horizon work — gpt-6-astra, OpenAI's most capable model.
• Classification, extraction, routing, bulk jobs — gpt-6-luna; set reasoning_effort to low or none to keep reasoning tokens down.

Call GPT-6 on Kunavo

Kunavo serves the GPT-6 models on the Responses API, which OpenAI builds them for, and on Chat Completions. Keep the OpenAI SDK and change base_url:

import os
from openai import OpenAI

client = OpenAI(
    api_key=os.environ["KUNAVO_API_KEY"],
    base_url="https://api.kunavo.com/v1",
)

# GPT-6 models reason before they answer; reasoning bills as output.
resp = client.responses.create(
    model="gpt-6-sol",
    reasoning={"effort": "medium"},
    input="Write a migration plan for splitting this monolith into services.",
)
print(resp.output_text)

# The same models on Chat Completions — swap the id for gpt-6-luna or
# gpt-6-astra and nothing else changes. reasoning_effort maps to
# the Responses API's reasoning.effort.
resp = client.chat.completions.create(
    model="gpt-6-luna",
    reasoning_effort="low",
    messages=[{"role": "user", "content": "Classify this ticket: 'refund not received'"}],
)
print(resp.choices[0].message.content)

For Codex, point a custom provider at Kunavo and set gpt-6-sol — the Codex CLI page has the config. Keys are created at /app/keys.

FAQ