minirouter.sh
Open navigation
Get a key

OpenAI

GPT-6 Astra

OpenAI's flagship for demanding coding, research, and document work. Inspect the evidence, compare the cost, and start with the exact API configuration below.

Get a MiniRouter keyCompare this modelSee the evidence
openai/gpt-6-astratext + image + pdf inputtext output

Replaces GPT 5.6 Sol: input $4.2 → $10.5 per 1M (+150%), output $21 → $52.5 per 1M (+150%), same context window.

Run this model

openai/gpt-6-astra

Use an endpoint supported by this model.

Get a key for this setup →
Thinking guide →
const response = await fetch('https://api.minirouter.sh/v1/responses', {  method: 'POST',  headers: {    Authorization: `Bearer ${process.env.MINIROUTER_KEY}`,    'Content-Type': 'application/json',  },  body: JSON.stringify({    model: 'openai/gpt-6-astra',    input: 'Why is the sky blue?',    max_output_tokens: 1024,  }),}) const data = await response.json()console.log(data.output)
Read docs →

Use Responses for Astra tool calls. Remove temperature, top_p, and logprob options. Start with a text request; client configurations remain subject to verification.

Model overview

Input / 1M
$10.5
All-in MiniRouter rate
Output / 1M
$52.5
All-in MiniRouter rate
Context
1.1M
1,050,000 tokens
Max output
128K
tokens
Reasoning
low → max
five effort settings
Tool calls
Responses
required by OpenAI
Announced
3 Sept 2026
publisher release
Publisher-reported results

Compare the work you need it to do.

OpenAI reports gains over GPT-5.6 Sol on terminal tasks, automation, and scientific reasoning. Scores are maxima across evaluated effort settings; its research/API harnesses may differ from your workload.

Astra and GPT-5.6 Sol

Terminal-Bench 4.0

Scores reproduced from OpenAI's launch comparison.

Reported score

0 to 100%

  • GPT-6 Astra57.9%
  • GPT-5.6 Sol37.3%
Source: OpenAI GPT-6 Astra announcement ↗. Publisher results; MiniRouter did not run these evaluations.
View chart data
Terminal-Bench 4.0 values
ModelScore
GPT-6 Astra57.9%
GPT-5.6 Sol37.3%

Astra and GPT-5.6 Sol

AutomationBench

Scores reproduced from OpenAI's launch comparison.

Reported score

0 to 100%

  • GPT-6 Astra41.4%
  • GPT-5.6 Sol18.1%
Source: OpenAI GPT-6 Astra announcement ↗. Publisher results; MiniRouter did not run these evaluations.
View chart data
AutomationBench values
ModelScore
GPT-6 Astra41.4%
GPT-5.6 Sol18.1%

Astra and GPT-5.6 Sol

GPQA Diamond

Scores reproduced from OpenAI's launch comparison.

Reported score

0 to 100%

  • GPT-6 Astra96.0%
  • GPT-5.6 Sol94.6%
Source: OpenAI GPT-6 Astra announcement ↗. Publisher results; MiniRouter did not run these evaluations.
View chart data
GPQA Diamond values
ModelScore
GPT-6 Astra96.0%
GPT-5.6 Sol94.6%
Artificial Analysis

A second view of the tradeoff.

The checked-in measurements are dated 2026-09-21. These evaluate the model's first-party API. They do not measure MiniRouter route speed or reliability.
All effort settings and measurements

GPT-6 Astra (max)

Intelligence Index
52.7#2 of 462 families
Coding Index
76.9
Output speed
63.2 tok/s
Source: Artificial Analysis ↗. Scores use this effort setting; they are separate from the publisher's launch table.
View chart data
Independent Astra results
MetricValue
Intelligence Index52.7
Coding Index76.9
Output speed63.2 tokens/s
Plan the workload

Token price is only part of task cost.

Compare total input, reasoning, output, and retries on representative tasks. The current MiniRouter rates above are the starting point; the figures here describe OpenAI's published Standard pricing rules.
OpenAI model and pricing documentation
  1. [ 01 ]

    Budget input and output together

    OpenAI lists $10 per million input tokens and $50 per million output tokens. Its context window is 1,050,000 tokens, with up to 922,000 input and 128,000 output.

  2. [ 02 ]

    Account for long prompts

    Above 272K input tokens, OpenAI applies 2× input and cache rates and 1.5× output rates to the full request. Cache writes and Fast processing have separate rates.

  3. [ 03 ]

    Check a representative request

    Use MiniRouter's request usage and cost records to measure your workload. Independent speed results and publisher task-cost estimates do not establish the cost or latency of your route.

Moving from GPT-5.6 Sol

Check three settings before switching.

The model ID is only one part of the change. Use these OpenAI-documented requirements when adapting your requests, and verify any tool or client workflow through MiniRouter separately.
OpenAI Astra migration guide
  1. [ 01 ]

    Choose a supported reasoning effort

    Use low, medium, high, xhigh, or max. Replace none or minimal with low, then evaluate the result on your own tasks.

  2. [ 02 ]

    Remove sampling overrides

    Astra does not accept temperature, top_p, or top_logprobs. Remove Chat Completions logprobs and Responses output-text logprob includes too.

  3. [ 03 ]

    Use Responses for tools

    Text Chat Completions is supported by OpenAI, but Astra tool calling requires Responses. OpenAI's async tools, computer use, and steering features need their own gateway support; a model listing alone does not establish that support.

Route

One route, standard clients.

Authenticate with a MiniRouter bearer key. Caller-provided upstream credentials are not used on this route.
Read model setup docs
Base URLhttps://api.minirouter.sh/v1
Model IDopenai/gpt-6-astra
Endpoints/v1/chat/completions/v1/messages/v1/responses
Parametersmax_tokens, stop, tools, tool_choice, reasoning, include_reasoning
Structured outputNot enforced; a requested format is dropped and the answer is plain text.
ModalitiesInput: text, image, pdf · Output: text

Configured providers. Priority follows the live catalog order.

ProviderPriorityInput / 1MOutput / 1MParametersAccess
vercel
1$10.5$52.5max_tokens, stop, tools, tool_choice, reasoning, include_reasoningMiniRouter balance

Every model, one key.

Get an API keyRead the docs