OpenAIGPT-6

GPT-6 Astra

GPT-6 Astra is OpenAI’s model for difficult reasoning and extended coding work. Its large context limit is useful for substantial inputs, but the price changes before that limit is reached.

A large context changes the cost of the whole request

Above 272,000 input tokens, Astra charges twice the input and cache rates and 1.5 times the output rate for the full request. The advertised context capacity and the base price therefore describe different operating conditions.GPT-6 Astra model reference

Image understanding is separate from image generation

Astra accepts image input. Image generation is a supported tool in the Responses API, rather than an image output modality of the language model itself.GPT-6 Astra model reference

Build a decision task before choosing the model

Practical guidance. For a large technical review, ask for a decision, its supporting file references and unresolved questions. Keep a few difficult examples to compare with a less costly model. A long response is not, by itself, evidence that the review is correct.

Specifications & access

GPT-6 Astra specifications
DetailValue and conditions
AccessAPI
Model identifiersgpt-6-astra
Context window1,050,000 tokens

API context limit

GPT-6 Astra model reference
Maximum output128,000 tokens

Maximum output

GPT-6 Astra model reference
Input price10 USD per million tokens

Standard text input; up to 272,000 input tokens

GPT-6 Astra model reference
Output price50 USD per million tokens

Standard text output; up to 272,000 input tokens

GPT-6 Astra model reference
Input and outputText and images in; text out. Audio and video are not supported.GPT-6 Astra model reference
Reasoning controlslow, medium, high, xhigh and maxGPT-6 Astra model reference
Fine-tuningNot supportedGPT-6 Astra model reference
Maximum input922,000 tokens within the total context window.GPT-6 Astra model reference

Related models

  • A model for coding, tool use and complex professional tasks.

    Launch date not confirmedAPI
  • A model for visual reasoning, screen interaction and writing with optional reasoning.

    Launch date not confirmedAPI