GPT-6 Astra

GPT-6 Astra handles coding and 3D scene code; tools build the assets. Access differs across Chat, Work, Codex, API and paid plans.

Platform: OpenAI

Language ModelReasoningCodingComputer UseAgent WorkflowsOpenAI

OpenAI-hosted / plan- and product-specific access

Standard: $10/1M input, $1/1M cached input, $12.50/1M cache write, and $50/1M output; above 272K input uses 2x input/cache and 1.5x output; Batch/Flex 50%, Fast 2x

Commercial

Visual examples

Side-by-side donut renders from the GetLLMs Codex and Blender test: direct modeling on the left and a reference-guided revision on the right. — Our iterative Codex/Blender workflow under matched camera and lighting. This is a tool-assisted example, not a native image-output claim or a controlled model benchmark.
GetLLMs hands-on test · September 6, 2026 · Open full-size image

🚀Function Overview

GPT-6 Astra is OpenAI's highest-capability route for hard end-to-end work, with long context and a broad hosted-tool surface.

Key Features

Official API model ID gpt-6-astra
Released September 3, 2026; GPT-6 Pro is available in Chat on eligible Pro, Business and Enterprise plans, while Plus includes Astra in Work and Codex
1,050,000-token context window and 128,000-token maximum output
Text and image input with text output
Generates scene code for editable Blender geometry; Blender and rendering tools produce the 3D files and images, while the model API returns text
Low, medium, high, xhigh, and max reasoning effort
Standard pricing of $10 input, $1 cached input, $12.50 cache write, and $50 output per million tokens
Requests above 272K input tokens use 2x input/cache and 1.5x output rates for the full request
Responses API tools include search, image generation, code interpreter, hosted shell, apply patch, skills, computer use, MCP, and tool search
April 30, 2026 knowledge cutoff
OpenAI's first model classified at the Critical cybersecurity capability threshold

Use Cases

•Build and revise editable Blender scenes and architectural prototypes through a tool-enabled environment
•Difficult end-to-end coding and repository work
•Research workflows combining search, files, tools, and long context
•Computer-use and document-creation agents
•Escalation after cheaper models fail task acceptance checks
•Approved high-capability defensive security evaluation

⚙️Input Parameters

model

string

Use gpt-6-astra when the active OpenAI organization and product surface have access.

input

object

Text or supported image content sent through the Responses API or another documented endpoint.

reasoning.effort

string

Choose low, medium, high, xhigh, or max based on task difficulty and cost tolerance.

tools

array

Optional OpenAI-hosted or developer-defined tools supported by the active Responses API contract.

💡Usage Examples

Example 1

Input Parameters

{
  "model": "gpt-6-astra",
  "input": "Inspect this repository, reproduce the failing workflow, implement the smallest fix, and verify the result.",
  "reasoning": {
    "effort": "high"
  }
}

Output Results

GPT-6 Astra returns a tool-using response when the account has access. Validate permissions, task success, token use, and review burden before production routing.

Quick Actions

Use NowView Documentation

Technical Specifications

Hardware Type: OpenAI-hosted / plan- and product-specific access
Commercial Use: Supported
Pricing: Standard: $10/1M input, $1/1M cached input, $12.50/1M cache write, and $50/1M output; above 272K input uses 2x input/cache and 1.5x output; Batch/Flex 50%, Fast 2x
Platform: OpenAI