GPT-6 Astra

Available

An OpenAI model with a 1.05M-token context window, image input and tool calling. On the OpenAI API since September 3, 2026.

Access details: OpenAI API

The model ID to use in API requests.

gpt-6-astra

Updated Oct 20, 2026

Sources & notes

Capabilities & limits

Context window: 1.05M
Maximum output: 128K
Image input: Supported
Tool calling: Supported
Structured output: Supported

Pricing

USD / 1M tokens

Input: $10.00
Output: $50.00

Standard rates per million tokens. Discounts and long-context rates, where offered, are on OpenAI's pricing page.

Full rate card 8 rates

All prices in USD.

Standard input

$10.00 per 1M tokens

standard processing
excludes cached input
excludes Batch API
excludes tool fees

Standard output

$50.00 per 1M tokens

standard processing
excludes Batch API
excludes tool fees

Cached input

$1.00 per 1M tokens

prompt caching

Cache writes

$12.50 per 1M tokens

1.25x uncached input rate

Long-context input

$20.00 per 1M tokens

request has more than 272K input tokens
rate applies to the full request

Long-context output

$75.00 per 1M tokens

request has more than 272K input tokens
rate applies to the full request

Long-context cached input

$2.00 per 1M tokens

request has more than 272K input tokens
rate applies to the full request

Long-context cache writes

$25.00 per 1M tokens

request has more than 272K input tokens
rate applies to the full request

Prompts over 272K input tokens: 2x input/cache and 1.5x output; Batch and Flex: 50% of Standard; Fast mode: 2x applicable rates

Price history

Sep 13, 2026

Input $10.00 per million tokens

Oct 20, 2026

Input $10.00 per million tokens

Sep 13, 2026

Output $50.00 per million tokens

Oct 20, 2026

Output $50.00 per million tokens

Technical details

Parameters: Not published
Active parameters: Not published
Architecture: Not published
Training / knowledge cutoff: Apr 30, 2026
Input modalities: Text, image
Output modalities: Text

From OpenAI

Official announcement ↗

How they describe it

Our most capable model, built for the hardest end-to-end work

Release & lifecycle

First available: Sep 3, 2026
available: As of Sep 13, 2026
available: As of Oct 20, 2026