Introduction to OpenAI's Latest AI Model: GPT-5.6
OpenAI, a leading company in artificial intelligence, has just launched its latest AI model called GPT-5.6. This model is claimed to be one of the most advanced innovations they have ever created. However, unlike its predecessors, GPT-5.6 is currently not accessible to the general public. OpenAI decided to release this model under a limited preview scheme.
This step was taken following a request from the White House to restrict the distribution of GPT-5.6 only to partners and certain organizations that have received approval from the U.S. government. This request arose because GPT-5.6's capabilities are considered comparable to Anthropic's latest AI model, Claude Mythos, which has also faced similar restrictions due to national security concerns.
GPT-5.6 Model Structure
GPT-5.6 has three model variants, each designed for specific purposes. Its flagship model, Sol, is intended for applications requiring high intelligence. Meanwhile, Terra is aimed at general use, and Luna is optimized for speed and efficiency. This division is similar to Anthropic's AI model family structure, which consists of Opus, Sonnet, and Haiku.
The top-tier model, Sol, offers superior capabilities across various fields, including programming, biology, and cybersecurity. With these abilities, Sol is expected to provide more effective and efficient solutions for various challenges faced by users in these areas.
Key Features of GPT-5.6
OpenAI also introduced two interesting new features for the Sol model. The first is Max Reasoning Effort, which allows the model more time to perform deeper reasoning. This feature is expected to improve the accuracy and quality of the model's responses.
The second feature is Ultra Mode, which enables GPT-5.6 to use multiple AI sub-agents to complete complex tasks more quickly. The combination of these two features is expected to bring higher efficiency and productivity in AI usage across various sectors.
In internal testing, GPT-5.6 Sol demonstrated impressive performance by scoring 91.9 percent on the TerminalBench 2.1 benchmark, a command-line-based coding test. This score is higher than previous models such as GPT-5.5 and Claude Mythos 5, indicating a significant improvement in the AI's programming capabilities.
Capabilities in Biology and Cybersecurity
In the field of biology, Sol is also recognized for delivering better genomic analysis results compared to GPT-5.5, with more efficient token usage. This shows that the new model is not only faster but also more economical in data processing.
In the cybersecurity sector, OpenAI claims that GPT-5.6 Sol can assist researchers in identifying and fixing security vulnerabilities more effectively. This is good news for professionals in security who are constantly seeking ways to protect systems from potential threats.
OpenAI has also added protective layers within the GPT-5.6 system to prevent misuse. This multi-layered security system is designed to protect the model from jailbreak techniques, which are attempts to manipulate AI to bypass existing security restrictions. Additionally, OpenAI conducts testing to handle various real-world attack scenarios to ensure the model remains safe to use.
With the launch of GPT-5.6, OpenAI demonstrates its commitment to continuous innovation and creating AI technology that is not only advanced but also safe to use. Although access is currently limited, many parties are eagerly anticipating the arrival of this model on the market to leverage its capabilities in various applications.
Source: https://tekno.kompas.com/read/2026/06/27/09590077/openai-rilis-gpt-5.6-model-ai-canggih-mirip-claude-mythos-
