OpenAI has launched a limited preview of its GPT-5.6 models, featuring three tiers aimed at coding, cybersecurity, and biology applications. This release comes shortly after US regulatory intervention and includes strengthened safeguards to address security concerns.
- Three-tier GPT-5.6 suite launched with specialized uses.
- Sol model priced competitively and enhanced for safety.
- Preview closely monitored under US government oversight.
What happened
OpenAI announced the limited preview of its GPT-5.6 model family, which consists of three variants: Sol as the flagship, Terra targeting medium-volume tasks, and Luna designed for affordability and speed. These models are optimized for sectors such as coding, cybersecurity, and biology and feature new operational modes for deeper reasoning and enhanced sub-agent capabilities.
The preview followed a request from the US Trump administration to stagger the model’s release for security evaluation. Pricing for Sol is set at $5 per million input tokens and $30 per million output tokens, undercutting competitors like Anthropic’s Claude Fable 5. The preview period involves extensive automated testing and third-party evaluation, expected to continue over the next two weeks.
Why it matters
This launch highlights OpenAI’s attempt to balance broad accessibility with the increasing need for robust safety measures amid growing regulatory scrutiny. The company emphasizes that GPT-5.6 Sol has the most comprehensive safety stack to date, designed to prevent misuse, especially in sensitive cyber-related contexts, including blocking illegal assistance and jailbreak attempts.
The involvement of the US government to vet customers during the preview period underscores the evolving relationship between AI developers and regulators. OpenAI’s cooperation represents a potential model for future AI governance frameworks, showing how industry and authorities might collaborate to manage risks associated with powerful AI systems.
What to watch next
The preview phase will provide critical insights into how well the enhanced safety layers perform in real-world scenarios, particularly in preventing harmful or unauthorized use. OpenAI’s continued collaboration with third-party testers and regulatory bodies will be key to assessing the readiness of GPT-5.6 models for general availability.