Roar AI
DiscordGet started
All models
OpenAI · Chat

GPT-6.2 Astra

GPT-6.2 Astra costs $8 per million input tokens and $40 per million output tokens on Roar AI. OpenAI’s newest flagship, strongest on science and maths, and slow to answer.

Input$8per 1M tokens
Output$40per 1M tokens
Context window1.05Mtokens

GPT-6.2 Astra API pricing

Input$8 per 1M tokens
Output$40 per 1M tokens
Cached input$0.40 per 1M tokens
Cache write$10 per 1M tokens
Batch input$4 per 1M tokens
Batch output$20 per 1M tokens
Input, long requests$16 per 1M tokens
Output, long requests$60 per 1M tokens

Above 272,000 input tokens, the whole request is billed at the long-request rates.

Example: 1,000 requests with 3,000 input and 1,000 output tokens each cost $64. Every rate is published. See every model.

Rates as of 20 November 2026.

Specs

Made byOpenAI
ReleasedNovember 2026
TypeChat
Context window1.05M tokens
Longest answer128K tokens
ReadsText, images
Thinks before answeringAlways on
WeightsClosed
Processed inGlobal
Model idgpt-6.2-astra

Is GPT-6.2 Astra worth the price?

For hard science and maths, yes, it is the one to test first. Vals AI ranked it first on its Science Index and on Terminal-Bench-Science.

For coding and everyday professional work, probably not. Artificial Analysis puts GPT-6.2 Sol level with the older GPT-6 Astra at a fraction of the cost per task, and Claude Sonnet 5.5 and Opus 5.5 both scored higher than Astra on the Vals Index. Try those first, and move up only if their answers fall short.

GPT-6.2 Astra and the models to compare it with, priced per 1M tokens
ModelInputOutputChoose it for
GPT-6.2 Astra$8$40Hard science and maths.
GPT-6.2 Sol$2$10Everyday work and agents, much faster.
Claude Sonnet 5.5$2$10Coding and documents. It scored higher on the Vals Index.
Claude Opus 5.5$4$20Long, judgement-heavy work.

Prices per 1M tokens.

About GPT-6.2 Astra

GPT-6.2 Astra arrived on 17 November 2026, eight weeks after GPT-6 Astra, and OpenAI calls it its most capable model for the hardest projects, from start to finish. Independent tests back that up: it scored 62 on the Artificial Analysis Intelligence Index at its highest effort, five points above GPT-6 Astra and three above Claude Opus 5.5.

Its clearest lead is in science. Vals AI ranked it first on its Science Index and first on Terminal-Bench-Science, a test of scientific work at the command line. On broader professional work it is less far ahead: on the Vals Index, which covers finance, legal and coding tasks, it scored below Claude Opus 5.5, Claude Sonnet 5.5 and Gemini 4 Argon.

We tested its cost and speed at each effort setting through our API on launch week: GPT-6.2 Astra, GPT-6.2 Sol and Gemini 4 Neon: pricing and API tests.

What is GPT-6.2 Astra good for?

  • Advanced science and maths problems.
  • Long research tasks where the quality of the answer matters more than the wait.
  • Operating software and working with documents across a long project.

GPT-6.2 Astra limitations

  • Slow. At its highest effort, Artificial Analysis measured over three minutes before the first word of the answer. Lower effort settings answer much sooner.
  • Artificial Analysis found its coding-agent score no better than GPT-6 Astra’s.
  • On Artificial Analysis’s AA-Omniscience test it gave a wrong answer rather than declining in 59% of the questions it could not answer correctly.
  • It always reasons; it cannot be told to answer straight away.

How to use the GPT-6.2 Astra API

Use the OpenAI SDK you already have: point it at Roar AI and set the model name. Read the docs.

from openai import OpenAI

client = OpenAI(base_url="https://api.roar-ai.com/v1", api_key="roar_live_…")

reply = client.chat.completions.create(
    model="gpt-6.2-astra",
    messages=[{"role": "user", "content": "Summarise this contract in plain English."}],
)

GPT-6.2 Astra FAQ

How much does GPT-6.2 Astra cost?

Input is $8 per 1M tokens; output is $40 per 1M tokens. For example, 1,000 requests with 3,000 input and 1,000 output tokens each cost $64.

Can I use GPT-6.2 Astra alongside other models?

Yes. Every model on Roar AI is on the same API key and the same monthly invoice, so switching from GPT-6.2 Astra to another model only means changing the model name in your code.

Is GPT-6.2 Astra open source?

No. OpenAI does not publish the weights, so it is only available through an API like this one.

What is the context window of GPT-6.2 Astra?

GPT-6.2 Astra can read up to 1.05M tokens in one request, and write an answer of up to 128K tokens.

Can GPT-6.2 Astra read images and PDFs?

It reads images directly. You can still send it a PDF: Roar AI turns each page into text first, at no extra charge.

Start using GPT-6.2 Astra

Create a free account, get an API key and call GPT-6.2 Astra with the OpenAI SDK.