OpenAI said it would open public access to the new line of GPT-5.6 models on Thursday. For several weeks, they were available only to a limited number of partners, since the release was an additional check by the American authorities. Now the company has received permission to launch a wider launch.
Increased attention to GPT-5.6 is due to the fact that the new generation of models copes noticeably better with finding vulnerabilities in the software. For developers, such opportunities help to quickly find errors in the code, but the same tools are able to use attackers to find new attack points.
After the completion of the audit, OpenAI announced that from July 10, models will be available to a wide audience. Together with the flagship GPT-5.6, the company will release lighter versions of Terra and Luna, designed for different tasks and levels of computational load.
According to OpenAI, the GPT-5.6 family received stronger logical analysis capabilities, it became better to write program code and work faster than previous models. The company also reports on enhanced security mechanisms that were developed in parallel with the new architecture.
The GPT-5.6 history continues a broader trend. In recent months, the US authorities have been paying increased attention to the most powerful AI models. Previously, for similar reasons, the restrictions were affected by the development of Anthropic: access to the most productive models, the company expanded only after the introduction of additional protection measures.
At the same time, a new approach to the release of the most powerful artificial intelligence systems is being formed in the United States. Companies can provide state structures with information about new models in advance to assess the possible risks before the start of mass distribution. GPT-5.6 was one of the first commercial models of OpenAI, the launch of which took place only after such a procedure.
Increased attention to GPT-5.6 is due to the fact that the new generation of models copes noticeably better with finding vulnerabilities in the software. For developers, such opportunities help to quickly find errors in the code, but the same tools are able to use attackers to find new attack points.
After the completion of the audit, OpenAI announced that from July 10, models will be available to a wide audience. Together with the flagship GPT-5.6, the company will release lighter versions of Terra and Luna, designed for different tasks and levels of computational load.
According to OpenAI, the GPT-5.6 family received stronger logical analysis capabilities, it became better to write program code and work faster than previous models. The company also reports on enhanced security mechanisms that were developed in parallel with the new architecture.
The GPT-5.6 history continues a broader trend. In recent months, the US authorities have been paying increased attention to the most powerful AI models. Previously, for similar reasons, the restrictions were affected by the development of Anthropic: access to the most productive models, the company expanded only after the introduction of additional protection measures.
At the same time, a new approach to the release of the most powerful artificial intelligence systems is being formed in the United States. Companies can provide state structures with information about new models in advance to assess the possible risks before the start of mass distribution. GPT-5.6 was one of the first commercial models of OpenAI, the launch of which took place only after such a procedure.
