OpenAI Previews GPT-5.6 Sol With Restricted Access

OpenAI released three versions of its GPT-5.6 model on Friday: Sol, Terra, and Luna. These models are being offered as a limited preview to a select group of companies, a move that includes ongoing engagement with the U.S. government. The Sol model is positioned as the latest flagship and most powerful iteration, while Terra aims to balance efficiency with performance. Luna is specifically fine-tuned for enhanced speed and affordability.

This limited release signifies a strategic approach by OpenAI to gather feedback and ensure robust performance and security before a wider public rollout. The inclusion of U.S. government entities in this preview suggests a focus on applications relevant to national security and public sector use cases. The company has emphasized stronger cybersecurity safeguards for these new models, indicating a proactive stance on addressing potential vulnerabilities and misuse.

While specific performance benchmarks for Sol, Terra, and Luna were not detailed in the announcement, the tiered release strategy suggests distinct capabilities and target applications for each model. Sol is expected to represent the cutting edge of OpenAI's AI development, likely showcasing advancements in reasoning, comprehension, and generation across various modalities. Terra's positioning indicates a focus on practical, widespread deployment where computational resources are a consideration.

Luna's emphasis on speed and affordability points towards applications requiring rapid response times or cost-effectiveness, such as real-time customer service or large-scale data processing. The preview phase is critical for OpenAI to refine these models based on real-world usage and feedback from its partners, particularly concerning their integration into existing systems and workflows. The company has not yet provided a timeline for general availability of GPT-5.6.