OpenAI

GPT-6.2 Astra

OpenAI's new Astra for demanding software, research and professional documents. Bring a detailed brief and a way to verify the result.

Successor to GPT-6 Astratext + image inputtext output

Replaces GPT-6 Astra: input $10.5 → $8.4 per 1M (-20%), output $52.5 → $42 per 1M (-20%), same context window.

Run this model

openai/gpt-6.2-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/chat/completions', {  method: 'POST',  headers: {    Authorization: `Bearer ${process.env.MINIROUTER_KEY}`,    'Content-Type': 'application/json',  },  body: JSON.stringify({    model: 'openai/gpt-6.2-astra',    messages: [{ role: 'user', content: 'Why is the sky blue?' }],    max_tokens: 1024,  }),}) const data = await response.json()console.log(data.choices[0].message.content)
Read docs →

Start with text and low reasoning effort. These models can reserve their full 128,000-token output allowance even when you request a short answer; keep enough balance for the hold. Omit temperature and top_p, and do not set reasoning to none.

Model overview

Input / 1M
$8.4
All-in MiniRouter rate
Output / 1M
$42
All-in MiniRouter rate
Context
1.1M
1,050,000 tokens
Max output
128K
tokens
Reasoning
low → max
Five effort settings
Best fit
Complex projects
Code, research and documents
Announced
17 Nov 2026
publisher release

Compare text-model alternatives or estimate token costs for your workload.

When the task is difficult

Use Astra where a stronger answer changes the outcome.

Choose a task with several constraints: reviewing a migration, reconciling conflicting documents or planning an implementation with measurable acceptance criteria. Keep Sol as a lower-cost baseline on the same task.
  1. [ 01 ]

    Complex code changes

    Supply the relevant interfaces, tests and failure symptoms. Ask for a plan that addresses the failure before proposing code.

  2. [ 02 ]

    Document synthesis

    Provide source labels, the intended audience and a concrete deliverable. Review every claim that affects a decision.

  3. [ 03 ]

    Research with checks

    Separate assumptions, observations and calculations. Ask for reproducible steps that another person can verify.

Start small

Make the first result easy to check.

Use a self-contained text task with a clear acceptance test before moving an existing workflow. Keep the exact model ID in your request so your comparison is reproducible.
  1. [ 01 ]

    Give it the evidence

    Provide the relevant code or extracted document text. Ask for source references where the answer depends on a particular passage.

  2. [ 02 ]

    Set the effort deliberately

    The catalog offers low, medium, high, xhigh and max. Start at low, then compare a higher setting only when the task needs it.

  3. [ 03 ]

    Check what came back

    Run proposed tests and verify calculations. Validate machine-readable answers in your application before using them.

Moving an existing workflow

Keep a baseline you can rerun.

Run the same tasks on the old and new model, with the same tools and acceptance checks. Compare successful results and total billed usage; token prices alone do not tell you the cost of a completed task.
  1. [ 01 ]

    Keep sampling options out

    Omit temperature and top_p. Reasoning cannot be disabled with none.

  2. [ 02 ]

    Separate the model from the application

    A model response does not run a browser, edit a file or execute a command. Your application supplies those tools and controls their permissions.

  3. [ 03 ]

    Read the current route details

    Check the model page before switching protocols or inputs. Schema-enforced output is not currently exposed for these MiniRouter routes.

Route

One route, standard clients.

Authenticate with a MiniRouter bearer key. Caller-provided upstream credentials are not used on this route.
Base URLhttps://api.minirouter.sh/v1
Model IDopenai/gpt-6.2-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 · Output: text

Configured providers. Priority follows the live catalog order.

ProviderPriorityInput / 1MOutput / 1MParametersAccess
vercel
1$8.4$42max_tokens, stop, tools, tool_choice, reasoning, include_reasoningMiniRouter balance

Every model, one key.