Key takeaways

• GPT-6 Astra is OpenAI's flagship model designed for demanding end-to-end tasks like advanced analysis, software engineering, deep research, scientific work, cybersecurity, and document creation.
• It excels particularly in long-horizon multi-step workflows and computer/browser use, indicating strong capabilities in complex, sequential tasks.
• The model utilizes a novel "recurrent depth" reasoning technique and an MoE (Mixture of Experts) architecture, trained on over 100,000 GPUs, suggesting a highly sophisticated and computationally intensive design.
• With a context length of 1,050,000 and max tokens of 100,000, GPT-6 Astra is built to handle extremely large inputs and generate substantial outputs, crucial for its intended complex applications.
• Released as a proprietary text+image+file-to-text model, its high cost per token (input: 1e-05, output: 5e-05) reflects its advanced capabilities and the extensive resources used for its development and deployment.

Technical specifications

Context length

1M tokens

Max output

—

Architecture

MoE

Modalities

text+image+file->text

Parameters

—

Open weights

No

Licence

Proprietary

Regions

USA

Vendor

OpenAI

Provider model key

openai/gpt-6-astra