OpenAI announced on Friday the launch of a limited preview of its new GPT-5.6 model series, aimed at enhancing performance in programming, cybersecurity, and biological research, while supporting complex long-duration tasks.
The rollout comes with a delay in the full public release, following a request from the US government for additional security reviews.
Advanced Capabilities in GPT-5.6
According to the company, the new series includes three primary models:
– Sol
– Terra
– Luna
The flagship model, Sol, is designed to handle advanced, multi-step tasks and deliver improved performance across technical domains.
OpenAI stated that the models include enhanced capabilities for:
– Code generation and debugging
– Vulnerability detection
– Scientific data analysis
– Complex reasoning tasks
Extensive Security Testing
OpenAI confirmed that the models underwent extensive security testing, totaling approximately 700,000 processing hours.
These tests were conducted in collaboration with external partners, who will continue evaluating the models during the preview phase to ensure safety and mitigate misuse risks.
The company described its new safety framework as the “strongest to date,” with improved mechanisms for identifying vulnerabilities and preventing harmful use.
Limited Access and Delayed Full Release
OpenAI said it has postponed the full public release of GPT-5.6 at the request of the US government.
For now, access is restricted to a small group of trusted partners, whose participation has been communicated to US authorities.
The company emphasized that this limitation is temporary and that broader availability is expected in the coming weeks.
US Concerns Over AI Risks
The delay reflects increasing scrutiny by US regulators over advanced AI systems, particularly concerns related to:
– Cybersecurity threats
– Potential misuse in cyberattacks
– Military or dual-use applications
OpenAI noted it is working with government agencies to develop a long-term framework for reviewing advanced AI models.
OpenAI Position on Government Access
Despite cooperating with authorities, OpenAI stated it does not believe early government access to models should become a permanent requirement.
The company warned that such a policy could limit access for developers, businesses, and cybersecurity experts worldwide.
Key Points
– OpenAI launches limited preview of GPT-5.6
– Three models: Sol, Terra, Luna
– Enhanced capabilities in coding and cybersecurity
– 700,000 hours of security testing
– Full release delayed at US government request
– Broader rollout expected in coming weeks
Conclusion
The release of GPT-5.6 marks another step forward in AI development, combining advanced capabilities with heightened security measures. The temporary delay highlights the growing balance between innovation and regulation as governments and companies navigate the risks and opportunities of next-generation AI systems.
