GPT-6 Astra
OpenAI flagship. Terminal-Bench 4.0 57.7%, DeepSWE 74.1%, GPQA Diamond 96%
- Vendor
- OpenAI
- License
- closed
- Released
- 2026-09-03
- Parameters
- Undisclosed
- Context window
- 1M tokens
Capabilities
Text, image, and file input with tool calling, extended reasoning, and computer use. OpenAI reports OSWorld 2.0 at 72.6% and GPQA Diamond at 96.0%.
- Vision — image input
- Audio — audio input (not confirmed)
- Tools — function calling / tool use
- Reasoning — extended thinking
- Computer use — desktop / browser control
Benchmarks
| SWE-Bench Verified | Not published |
|---|---|
| SWE-Bench Pro | Not published |
| Terminal-Bench | Not published |
| Reasoning composite | Not published |
The reasoning composite is a directional 0–100 score aggregated across public reasoning evals — comparable within this matrix, not an official vendor number.
Pricing
$10/M input tokens · $50/M output tokens.