GPT-6 Astra API: Pricing, the 272K Cliff and Setup

GPT-6 Astra is OpenAI's current flagship, live as model ID gpt-6-astra. The list price is $10 per 1M input tokens and $50 per 1M output tokens for prompts at or below 272,000 input tokens. Cross that line and the whole request bills at $20 input and $75 output per 1M.

That cliff is the number most pricing tables bury. It is not a marginal rate.

GPT-6 Astra API pricing #

All figures are USD per 1M tokens. The first row is the standard rate at or below 272K input tokens; the second is long-context pricing above 272K.

OpenAI also publishes $1 per 1M cached input and $12.50 per 1M cache writes. RouterPlex bills cached tokens at the full $10 input rate on this route, because the upstream does not pass the cache discount through. Price a cache-heavy agent against $10, not $1.

The 272K cliff applies to the whole request #

OpenAI's rule is the same shape as xAI's 200K cliff: once input tokens go past 272,000, the higher rate applies to every token in that request. There is no blended calculation.

271,000-token prompt on gpt-6-astra  →  271,000 × $10.00/1M  = $2.71
273,000-token prompt on gpt-6-astra  →  273,000 × $20.00/1M  = $5.46

Two thousand extra input tokens roughly double the input bill, and output on that same request lifts from $50 to $75 per 1M. If you are anywhere near 272K, trimming the prompt below the threshold is worth more than any other optimisation available to you.

How GPT-6 Astra compares on price #

Flagship list prices, USD per 1M tokens, on a 40,000-token prompt returning 2,000 tokens with no cache:

Astra is priced as a premium flagship, not a value one. It matches Claude Fable 5 on list rates and costs about twice a GPT-5.6 Sol turn. Pay that only when the newer model actually moves the task.

Live pricing: GPT-6 Astra, GPT-5.6 Sol, Claude Fable 5.

What OpenAI actually documents #

Reasoning tokens are billed as output at $50 per 1M ($75 above the cliff). A short reply is not necessarily a cheap one. Check usage.completion_tokens_details.reasoning_tokens if a simple prompt costs more than you expected.

OpenAI's Batch and Flex endpoints are 50% of Standard, and Fast is 2×. RouterPlex bills the standard list rate in the table above. Do not assume a Batch or Flex discount on this key.

Call GPT-6 Astra with RouterPlex #

Create a RouterPlex key, give it a hard budget, and send a standard chat-completions request:

curl https://api.routerplex.com/v1/chat/completions \
  -H "Authorization: Bearer $ROUTERPLEX_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-6-astra",
    "messages": [
      {"role": "user", "content": "Name the riskiest assumption in this migration plan."}
    ]
  }'

The Python version uses the regular OpenAI client — only the base URL, key and model ID change:

import os
from openai import OpenAI
 
client = OpenAI(
    api_key=os.environ["ROUTERPLEX_API_KEY"],
    base_url="https://api.routerplex.com/v1",
)
 
response = client.chat.completions.create(
    model="gpt-6-astra",
    messages=[{"role": "user", "content": "Design a rollback-safe deployment plan."}],
)
 
print(response.choices[0].message.content)

Claude Code and the Anthropic SDK reach the same model through the Anthropic-compatible format — set ANTHROPIC_BASE_URL=https://api.routerplex.com and use gpt-6-astra as the model. See the Claude Code setup guide for the full configuration.

Should you switch from GPT-5.6 Sol? #

• Hardest work, newest model — architecture reviews, high-risk migrations, long-horizon agents that need the 922K input window. Astra is the current OpenAI flagship. Budget $10 / $50 and stay under 272K unless you mean to pay the cliff.
• Same family, lower rate — keep GPT-5.6 Sol at $5 / $30 when evaluation does not show a quality gain that pays for doubling input.
• Anything near 272K — trim first. Two thousand extra tokens can double the request.

Astra is live on RouterPlex at OpenAI list input and output rates with no per-token markup, on the same prepaid key as Sol, Terra, Luna and the rest of the OpenAI catalog. See the live GPT-6 Astra price page before you commit a budget.